VENTED GAS EXPLOSIONS

By
Rafiziana Md. Kasmani

Submittedin accordance
with the requirementsfor the degreeof Doctorof
Philosophy

University of Leeds
Schoolof Process,EnviromnentalandMaterialsEngineering

Underthe supervisionof

ProfessorG.E.Andrews,PhD,B.Sc,C.Eng,M. I.MechE,SMAIAA, MASME,

MSAE

and

Dr.H.N.Phylaktou,Phd,B.Eng

September2008

The candidateconfirmsthat the work submittedis her own andthat appropriate

credit hasbeengivenwherereferencehasbeenmadeto the work of others

This copyhasbeensuppliedon the understandingthat it is

copyrightmaterialandthat
no quotationfrom the thesismay be publishedwithout properacknowledgement.
 i

ABSTRACT

Explosion venting technology is widely acceptedas the effective constructional

protectionmeasuresagainstgasand dust The
explosions. key problem in venting is
the appropriatedesign of the vent area necessaryfor an effective release of the
i.
material e. the developed
pressure during did
explosion not cause any damage to the
plant protected.Currentgasexplosion vent designstandardsin the USA (NFPA68,
2002) and European(2007)rely on the vent correlationfirst publishedby Bartknecht
in 1993 (Siwek, 1996).NFPA 68 also recommendsthe correlationof Swift (Swift,
1983)at low overpressures. For a vent to give no increasein overpressureother than
that due to the pressuredifferencecreatedby the massflow of unburntgasesthrough
the vent, the vent massflow rate is assumedto be equal to the maximum mass
burning rate of the flame and this considerationshould be used as the design mass
flow through the vent.Two different methods( Method I and Method 2) have been
proposedbasedon the S,,and S,,(E-1) to describethe maximum massburning rate
given as,

mb
= AS. p,, = CdeA,(2p,,Pd)o"
(1)

m,, =A., S, P,, =AS,, (E-I)P,, (2)

=Cd&4, (2p,, Po-ed)0'5

The equationgiven in (2) is slightly different from (1) as is about 6.5 times the mass
flow of the first methodasit takesthe effect of (E-1) whereE is the expansionratio.

A critical review werecarriedout for the applicability, validity and limitation on the
venting correlationsadoptedin NFPA 68 and EuropeanStandardwith 470 literature
experimentaldata,coveringa wide rangeof valuesfor vesselvolumeand geometries,
bursting vent pressure,P,, UD ratio, maximum reducedpressure,P"d and ignition
location. The fuels involved are methane,propane,hydrogen, town gas, ethylene,

acetone/air mixtures with the most hazardous near-stoichiornetric fuel-air

concentration.Besides,Molkov's equation (Molkov, 2001) which is regarded as
alternativeventingdesignoffered in NFPA 68 and Bradley and Mitcheson'sequation
for safeventing designwere also analysedon the experimentaldatafor their validity
andlimitation as well as the proposedmethods.
 ii

From the results,it is clear that Bartknecht'sequationgavea satisfactoryresult with

experimentaldata for K <-5 and Swift's equation(Swift, 1983) can be extendedto

wider rangefor Pred
> 200 the
mbar,providing parameter P, is addedinto the equation.
Method 2 gave a good agreementto most of the experimentaldata as it followed
for
assumptionsapplied correlationsgiven by Bradley and Mitchesonfor safeventing
design (Bradley and Mitcheson, 1978a,Bradley and Mitcheson, 1978b).It is also
proventhat the vent coefficient,K is confident to be usedin quantifying the vessel's
for the
geometry cubic vesseland useof As/Av term is morefavourable for non-cubic
vessels.

To justify the validity and applicability of the proposedmethods,seriesof simply

ventedexperimentswere carriedout, involving two different cylindrical volumes i. e.
0.2 and 0.0065 M3. It is found that self accelerationplays important role in bigger
vesselin determiningthe final P,,,. insidethe vessel.Method 2 gavecloserprediction
on P.. in respectwith otherstudiedcorrelations.

The investigationof ventedgasexplosionis exploredfurther with the relief pipe been

connected to the vesselat different fuel/air equivalenceratios, ignition position and P,
The resultsdemonstratethat the magnitudeof P,,. wasincreasedcorrespondingto the
increaseof Pv- From the experiments,it is found that peak pressurewith strong
acousticbehaviouris observedrelatedto increasein Pvand in somecases,significant
detonationspike was also observed,particularly in high burning velocity mixtures. It
is found that substantialamountof unburnt gasesleft inside the vesselafter the vent
burst is the leading factor in increaseof P,,, for high burning velocity mixtures at
centrally ignited. The associategasvelocitiesahead of the flame createhigh unburnt
gas flows conditions at entry to the vent and this give rise to high back pressures
which leadto the severityin final P.. insidethe vessel.

It was observedthat end ignition leads to a higher explosion severity than central
ignition in most cases,implying that central ignition is not a worst-casescenario in
gasventedexplosionsas reportedpreviously
 iii

CONTENTS

Abstract
Contents
List of figures vii
List of tables xii
Nomenclature xiii
Acknowledgements, xv
Publications by author xvi

CHAPTER 1- VENTED GAS EXPLOSION RESEARCH

-THE

1.0 Introduction I
.......................................................................................................
1.1 Venting gasexplosions 2
......................................................................................
1.1.1 Generaloverview 2
.......................................................................................
1.2 Overviewof the project studies 8
..........................................................................
1.2.1 Simply venting gasexplosions 9
...................................................................
1.2.1.1 Self-accelerationof sphericalflames 12
....................................................
1.2.1.2 Implication of the objectivesto the presentstudy 14
..................................
1.2.2 Ventedgasexplosionwith relief pipe attached 14
. ........................................
1.2.2.1 The influenceof ductpipe diameter 17
......................................................
1.2.2.2 Implication of the objectivesto the presentstudy 17
..................................

CHAPTER 2-EXPERIMENTAL SET-UP AND MEASURE'MENT TECNIQUES

2.1 Test facility 19

......................................................................................................
2.2 Apparatusdesignconsiderations 19
......................................................................
2.2.1 Test vessels 21
..............................................................................................
2.2.2 Dump vessel 23
.............................................................................................
2.3 Constructiondetails 23
.........................................................................................
2.3.1 Test vessels 23
..............................................................................................
2.3.2 Dump vessel 29
.............................................................................................
2.4 Vent coverselection 33
.........................................................................................
 iv

2.5 Instrumentationanddatacollection 34
..................................................................
2.5.1 Instrumentationtechniques 34
.......................................................................
2.5.1.1 Explosion pressure history 35
....................................................................
2.5.1.2 Measurement of flame positions 35
...........................................................
2.5.1.3 Thermocouples 36
.....................................................................................
2.5.1.4 Flame speedcalculation 37
.......................................................................
2.5.2 Data acquisition 39
.......................................................................................
2.6 Identification hazardsand safety measures 39
.......................................................
2.6.1 Vessel failure 40
..........................................................................................
2.6.2 Transmission of the explosion to auxiliary equipment 40
..............................
2.6.3 Releaseof combustible gas into the Test room 41
.........................................
2.7 Operating procedures 41
.......................................................................................
2.7.1 Fuel entrainment 41
......................................................................................
2.7.2 Ignition procedure 42
....................................................................................
2.7.3 Vessel purging 42
.........................................................................................
2.8 Ancillary equipment 42
........................................................................................
2.8.1 Isolation valves and pipe works 43
................................................................
2.8.2 Mixture pressure monitoring system 43
........................................................
2.8.3 Vacuum gate valve 47
...................................................................................
2.8.4 Evacuation system 47
....................................................................................
2.8.4.1 Vacuum pump A 47
..................................................................................
2.8.4.2 Vacuum pump B 47
..................................................................................
2.8.5 Ignition system 48
........................................................... I...I.,.......................
2.8.6 Nitrogen purging system 48
..........................................................................

CHAPTER 3-VENTED GAS EXPLOSIONS: THEORY AND CRITICAL REVIEW

3.0 Introduction 50
.....................................................................................................
3.1 Parameterinvolved in empirical equations 51
.......................................................
3.2 Venting mechanism 53
. ........................................................................................
3.3 Venting theory 58
.................................................................................................
3.4 Comparisonof Bartknecht and Swift correlationsas been adoptedin NFPA 68
usingderivedmethodson the samebasis.....................................................................
71
3.5 Reviewof the publishedexperimentaldata 74
. .....................................................
 V

3.5.1 Comparison of Bartknecht and Swift's correlations with published

experimental data 76
. ...................................................................................................
3.5.2 The influence of turbulent enhancementfactor, 0 80
.....................................
3.5.3 The influence of the vessel's geometry 85
.....................................................
3.5.4 Comparison between K and A, /A, term for vessel's geometries ............... 90
3.5.5 Evidence for an additional influence of vessel volume on Pred.................. 93

3.6 Concludingremarks 96
.........................................................................................

CHAPTER 4-SIMPLY VENTED GAS EXPLOSION: THE PRACTICAL

4.0 Introduction 98
.....................................................................................................
4.1 Generalfeaturesof experimentaltests 99
..............................................................
4.2 Generalexplosiondevelopment 102
.....................................................................
4.2.1 Resultsand discussionson Test vesselI 104
................................................
4.2.1.1 Maximum pressure,P.. as a function of equivalenceratio ................104
4.2.1.2 Flamespeeds 108
.....................................................................................
4.2.2 Resultsand discussionon Test vessel2 116
..................................................
4.2.2.1 Maximum pressure,Prn. 116
....................................................................
4.2.2.2 Flamespeeds 119
......................................................................................
4.3 Deflagrationto detonationin test vessel 122
.........................................................
4.4 Influenceof vent coefficient,K, volume andburstvent pressure,P, on Prnfoc.. 126
4.5 Comparisonbetweentheoryand experimentaldata 134
........................................
4.6 Concludingremarks 138
.......................................................................................

CHAPTER 5-VENTED DUCT GAS EXPLOSIONS

5.0 Introduction 140

..................................................................................................
5.1 Phenomenology of vesselventedthrough the duct......................................... 143
5.1.1 Effect of the relief duct fitted to the test vessel 145
.......................................
5.1.2 The influenceof staticburstingpressure,P, on maximumpressure,P.,,,,,149
5.1.3 The influenceof ignition position 153
...........................................................
5.1.4 The accelerationof flame towardsthe vent duct and associatedvent velocities
157
I...........I...I......I.........I............................................................................
5.1.5 The duct flame speedsand gasvelocities 176
................................................
 vi

5.2 The influenceof equivalenceratio on Pmax 179

.......................................................
5.3 Comparisonwith otherexperimentaldata 183
........................................................
5.4 Concludingremarks 185
..........................................................................................

CHAPTER 6- VENTED DUCT GAS EXPLOSIONS WITH BIGGER DIAMETER

PIPE ATTACHED

6.0 Introduction 187

.......................................................................................................
6.1 Effect of ductdiameteron Pmax 189
.........................................................................
6.2 Flame speedsanalysis 203
.......................................................................................
6.3 Pressurelossandunburntgasvelocity 207
.............................................................
6.4 Concludingremarks 212
..........................................................................................

CHAPTER 7- CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE

WORKS

7.0 Summaryof major findings 214

..............................................................................
7.1. Generaleffectsof ventingexplosions 214
...............................................................
7.1.1 Comparisonof theory andexperimentalresults 215
.......................................
7.1.2 Duct ventedgasexplosion 216
........................................................................
7.2 Recommendation for the future work 218
...............................................................

REFERENCES 221
....................................................................................

APPENDIX 230
........................................................................................
 vii

LIST OF FIGURES

Figure 1.1 Example of commercial vent covers and its application 12

Figure 2.1 Schematic picture of test facility and control room 20
Figure 2.2 Schematic drawing for Test vessel I and 2 26
Figure 2.3 Scaled drawing of Test vessel 3 with pipe attached 27
Figure 2.3a Scaled drawing of pipe section for Test vessel 3 28
Figure 2.3b Scaled drawing of pipe section for Test vessel 3 29
Figure 2.4 Dump vessel schematic designs 32
Figure 2.5Vent covers used in the tests. From top left: Melinex membrane,

aluminiurn foil. From bottom left: 100g A4 paper and magazine paper 34
Figure 2.6 Typical thermocouple traces indicating the change in voltage

output due to flame arrival 36

Figure 2.7 Exposedjunction, mineral insulated type-K then-nocouplesinside

the rig. 37
Figure2.8 Flow diagramillustratingsafetyandoperationalprocedure 44
Figure2.9 Exampleof operatingprocedurefor carryingexplosiontests 46
Figure3.1 Pressurebehaviourversustime for a closedand ventedexplosions 53
Figure3.2 The pressure-timevariationin a ventedexplosion.Graph

reproducedfrom Cooperet al (Cooper,FairweatherandTite, 1986) 57

Figure3.3 Bartknechtdatafor propane-airat different volumesat E=I OJ,
P, = 100mbars,(Bartknecht,1993).* is for methane-airat 30m3
on I/K. Reproducedfrom Andrews(Andrews,2004) 60
Figure3.4 Methane/airexperimentaldataas a function of K 77
Figure3.5 Propane/airexperimentaldataasa functionof K 78
Figure3.6 Hydrogen/airexperimentaldataas a function of K. 78
Figure3.7 Town's gasexperimentaldataas a functionof K 79
Figure3.8 Methanelairat variousgeometrieswithout 81
Figure3.9 Propane-airin variousgeometrieswithout 81
Figure3.10Methane/airwith turbulentfactor, 0 84
Figure3.11Town gas/airwith turbulentfactor, 84
Figure3.12Propane/airwith turbulentfactor, P 84
Figure3,13 Influenceof vessel'sshapeon methane/airexplosion o Sphereand
 viii

o cubic 87
Figure 3.14The influenceof rectangularvesselson Predfor propane/air
0 rectangular 89
Figure 3.15The influenceof variousgeometrieson Predfor town gas/air
0 rectangular,o cubic, o sphere 89
Figure 3.16The influenceof the vessel'sShapeOnPred-
A cylinder c,sphere 90
Figure 3.17Methane/airfor Predasa function of K 91
Figure 3.18Pmdof methane/airat variousgeometriesasa function of AdAv 92
Figure 3.19Ko asa function of vesselsizefor propane/air(4%) measured
at an initial pressure of I atm (Chippett, 1984) 94
Figure 3.20 Effect Of Predon vessel volume at constant K 95
Figure4.1 Typicalpressure-timehistoriesfor centreandendignition
in TestvesselI 103
Figure4.2 Typicalpressure-timehistoriesandflame positionin Test vessel2 103
Figure 4.3 Methane/airat different equivalenceratio, 0 105
Figure4.4 Propane/airat different equivalenceratio, 0 106
Figure4.5 Ethylene/airat different equivalenceratio, 0 106
Figure4.6 Maximumrateof pressurerise,dP/dt(max) at variousequivalence
ratios for propane/air(above)andethylene/air(below) 107
Figure 4.7 Hydrogen/airat differentequivalenceratio, 4) 107
Figure 4,8 Flamespeedsof methanelairat endandcentralignition as

a functionof equivalenceratio 109

Figure4.9 Flamespeedas a function of flame distancefrom the sparkfor open

venting 110
Figure 4.10 Flamespeedsof propane/airat end andcentralignition as

a functionof equivalenceratio 112

Figure 4.11 Time of flame arrival at maximumpressure,whenleaving
the vesselandat the comer regionasa function of propane/air
equivalenceratio 113
Figure 4.12 Diagramon the flame movementat end ignition and the
unburntgasesleft inside the vessel 113
Figure4.13 Flamespeedsof ethylene/airat endandcentralignition as

a function of equivalenceratio 114

 ix

Figure4.14 Flamespeedsof hydrogen/airat endandcentralignition as

a function of equivalenceratio 115
Figure4.15 Time of flame arrival at maximumpressure,whenleaving
the vesselandat the comer regionasa functionof equivalenceratio 116
Figure4.16 Pressure-time
history for differentfuelsat (D= 1.0
Figure4.17 Pressuretime history for maximumpressureinsidethe vessel(PI)
andpressureinsidethe dumpvessel(P6)at stoichiometrichydrogen/air 117
Figure4.18 MaximumpressurePn,,,for differentfuel/air mixturesas
),
a function of equivalenceratio 118
Figure4.19 Flamespeedsfor the studiedgas/airasa function of equivalenceratio.
Lewis no, Le andMarksteinno, Ma takenfrom this sources
(Clark andSmoot,1985,Searbyand Quinard,1990,Tseng,
Ismail andFaeth,1993) 121
Figure4.20 Maximumpressureswith andwithout spiketracesin hydrogenlair
explosionin Test vessel2 123
Figure4.21 Hydrogen/airexplosionat (D= 0.54 (16 % concentration)for
Test vessel1 at end andcentreignition 124
Figure4.22 Ethylene/airexplosionat (D= 1.0(6.5 % concentration)for
Test vesselI at endand centreignition 124
Figure4.23 Methane/airat different K for (D= 1.05 127
Figure4.24 The influenceof K on Pn andpredictedflame area,Af for

methane/airmixture 128
Figure4.25 Hydrogen/airat stoichiometricconcentrationfor different K 129
Figure4.26 Influenceof P,,on P for Testvessel2 133
......
historiesfor simply ventedandduct vented
Figure5.1 Pressure-time
explosionfor methane/airat 0=1.06 for endignition (initially open
venting).dP/dttraceswere given for both cases 146
Figure 5.2 Pressurerecords at selected positions along the test vessel for

methane/air at 4) = 1.06 for end ignition. AP2-3is the pressure difference

at the vent entrance, AP3-5is pressure difference inside the pipe and APS-6
is the duct exit pressure loss. Ti,, and T,,,,,indicate the time flame enters

and leavesthe duct respectively 149

Figure5.3 P... v P, on stoichiomeqicpropane/airand methane/airfor Im
lengthduct 150
 x

Figure5.4 P,,,,,v P, for ethylene/airat 0=0.8 andhydrogen/airat (D= 0.54 152

Figure5.4 Methane/airat P, 178mbar 155
Figure5.5 Propane/airat P, 178mbar 156
Figure5.6 Hydrogen/airat P, = 178mbar 156
Figure5.7 Ethylene/airat P, = 424 mbar 156
Figure5.8 Reproducedfrom Cooperet al(Cooper,M.G, Fairweather,M and
Tite, J.P, 1986).Flamepropagationmechanismat thevariousphases
of a ventedexplosion 162
Figure5.9 Flamespeeda function of the flame distancefrom the sparkfor
P, =0 for methane/airandpropanelair.(D= 1.0for both gas/mixtures 160
Figure5.10Flamespeedasa function of distancefrom the sparkfor
Pv=0 for ethylene/air(0 = 0.8) and hydrogen/air(0 = 0.54).
Notedthat only hydrogenline for NFPA 68 wasdrawnasthereis

no availabledatafor ethylenein the guidance 165

Figure5.11Hydrogen/airat (V= 0.54 for end ignition. P, = 209 mbar 167
Figure5.12Propane/airat (D= 1.0for end ignition. P, = 178mbar 168
Figure5.13Propane/airat 0=1.0 for endignition. P, = 424 mbar.Spiky pressure
peakis observedin this test 168
Figure5.14Methane/airfor 0=1.0 at centralignition. Pv= 424 mbar 169
Figure5.15Influenceof P, on flame speedupstreamof the vent for methane/air
(0 = 1.0)andpropane/air(0=1.0) with endandcentralignition 169
Figure5.16 Influenceof Pvon flame speedupstreamof the ventfor ethylene/air
(0 = 0.8) andhydrogen/air(0 = 0.54) with end andcentralignition 171
Figure5.17Methane/airfor endand centralignition as a functionof equivalence

ratio 172
Figure5.18 Propane/airfor end andcentralignition asa functionof equivalence
ratio 174
Figure5.19Ethylene/airfor endand centreignition asa function of equivalence

ratio 175
Figure5.20 Hydrogen/airfor endand centreignition as a function of equivalence
ratio 176
Figure5.21 Flamespeedin the duct pipe for propane/airasa function of
equivalenceratio at endignition 178
 xi

Figure5.22Flamespeedin the ductpipe for hydrogen/airasa functionof

equivalenceratio at endignition 179
Figure 5.23Methane/airat end(left) andcentre(right) ignition at variousP, 181
Figure 5.24Propanelairfor end (left) andcentre(right) ignition at variousP, 181
Figure 5.25Ethylene/airfor end(left) andcentre(right) ignition at variousP, 182
Figure5.26Hydrogen/airat end(left) andcentre(right) ignition at variousPv 182
Figure 5.27Comparisonbetweenotherpublishedexperimentaldataand
currentproject.The datareportedwere 10 % 4.5%
methane/air, propane/air,
5.2%ethylene/airand 16 % hydrogen/airand 5% acetone/air.
NFPA 68 correlation(NFPA68,2002)andBartknecht'sequation
(Bartknecht,1993)areusedfor 0< Ld >3m 184
Figure 5.28Predasa function of L/D 185
Figure 6.1 Pressuretime profile for methane/airat 0=1.08 190
Figure 6.2 Pressuretime profile for hydrogen/airat (D= 0.54 190
Figure 6.3 Pressuretime profile for hydrogen/airat end ignition. (D= 0.54 191
Figure 6.4 Pm of methane/airasa function of equivalenceratio 192
Figure 6.5 Pmax
of propane/airasa function of equivalenceratio 192
Figure 6.6 Pmax
of ethylene/airas a function of equivalenceratio 194
Figure 6.7 Pma,of hydrogen/airasa function of equivalenceratio 195
Figure 6.8 Time of flame arrival asa function of equivalenceratio 196
Figure 6.9 Flow interactionsfor methane/airat end ignition. (D= 1.08 197
Figure 6.10Flow regimesin vesselbeforeand after explosion-likecombustion
in duct,effectof the flame front distortion; a) narrowducts
b) largeduct.Reproducedfrom (Ponizy,B and Leyer,J.C, 1999b) 199
Figure 6.11Pressuretime historiesandflame arrival for ethylene/airat (D= 0.8 200
Figure 6.12Pressureinsidethe duct for ethylene/airas a function of equivalence
ratio 202
Figure 6.13Pressuretime historiesandflame arrival for propane/airat
(D= 0.8 for end ignition 203
Figure 6.14Averageflame speedsmeasured(a) in secondhalf of the vessel
(betweenT, andT3 as in Fig.2.3b)and (b) in the vent duct
(betweenT4andT9) as a function of equivalenceratio for methane/air 203
Figure 6.15Flamespeedsmeasureda) in secondhalf of the vessel
(betweenT, andT3 asin Fig.2.3b) and b) inside the duct
 xii

(betweenT4andT9)asa function of equivalenceratio for hydrogen/air 205

Figure6.16 Schlierencinephotographsof cellular flame developmentduring
explosionof H2/airmixture, (D= 0.26.Reproducedfrom
BradleyandHarper(Bradley,D andHarper,C.M, 1994) 208
 xiii

LIST OF TABLES

Table2.1 Test vesseldesigndetails 24

Table2.2 Dump vesseldesigndetails 30

Table2.3 List of Pt,,t of ventcoversused 33

Table2.4 Procedurein calculatingthe flame speedsfrom raw data 38

Table3.1 Studiedcorrelationsfor ventingof gasexplosions 71

Table3.2 Comparisonconstantvaluederivedfor Method I and 2 with respect

to Swift's C, fuel characteristicconstant 72

Table3.3 List of predictedP with respectwith Bartknecht'sC3constant 73

Table3.4 Ko valueusedfor vent areacalculation.All datawere excerpted

from BradleyandMitheson(BradleyandMitcheson,1978b)except

for acetone-airdata(Molkov, Dobashi,Suzukiand Hirano, 2000) 75

Table3.5 Summaryon averageturbulentfactor, P for different vessel'sshape

andfuels 83

Table4.1 Summaryof the experimentaltests 101

Table4.2 Summaryof different gas/airpropertiesat highestflame speeds.

Valuesof Table4.3 Time of flame arrival for Test vessel1 122

Table4.3 Time of flame arrival for Test vesselI 125

Table4.4 Summaryof experimentalPm. for Test vesselI and 2 for

K= 16.4.The ignition position is endignition 130

Table4.5 Summaryof experimentalaverageflame speed,Sr,,vgfor Test vesselI

and2 for K= 16.4.The ignition position is end ignition 132

Table4.6 Experimentaldataandcalculatedequationsfor testvesselsat (D= 1.0 136

Table4.7 Percenterror deviationfor experimentalandcalculationresult 136

 xiv

Table4.8 Comparisonof methane/airmixturewith differentP, for Test vessel2 137

Table 5.1 Explosiondatafor 4.0%propane/air,9.5%methane/air,5.2%

ethylene/airand 16%hydrogen/airmixtures.Flamespeed,Sf in the vessel
refersto the speedfrom the sparktowardsthe vent (TV4-Tv2)
while flame
in theduct is the averagevalueobtainedalong theduct (T.,,t-To).
St is the turbulentburningvelocity asSf/EwhereE is the adiabatic
expansionratio. KGis (dP/dt)V1/3and is basedon the maximum dP/dt
from pressuretransducerPI, which occursafter the flame hasexited the
duct.Kr,,, t is basedon therateof pressurerisejust prior to the flame entering
the ductpipe and is proportionalto the massburningrateupstreamof the
vent.* Spikingpressuretrace 163
Table 5.2 Predictedflame radiuswhenthe vent burstfor gas/air 166

Table 6.1 Summaryof experimentalresults(highestPmax

eitherend or

centralignition) with predictedvaluesfrom Eq.5.1,5.2,5.3 and 5.4

where Predwithout duct obtained from the simply vented experimental

data 199

Table6.2 Calculatedunburntgasvelocity at the duct entrance,insidethe duct pipe

andat the ductpipe exit for studiedfuel/air mixtures .210

 xv

NOMENCLATURE

Aý Total surface area of the vessel, M2

A, Vent area, M2
BR blockage ratio

Brt turbulent Bradley number

Cd orifice discharge coefficient
C"j speed of sound, m/s
dP/dt maximum rate of pressure rise
E expansion ratio
K Vent coefficient (= V2/3/A, )

Ka Karlovitz number

KG mixture reactivity index, bar-m/s

IJD length to diameter ratio
Le Lewis number
Ma Markstein number
Mach Mach number

Mb mass rate of burning

P"d reduced maximum pressure, bar
Pmax maximum overpressure,bar
Pr Prandtl number
P, vent bursting pressure, bar

R gas constant (82.0552 x 10-6M3bar/Kmol)

Re Reynolds number
Sf flame speed,m/s
S9 unburnt gas velocity, m/s
S" lan-iinar burning velocity, Ws
ST turbulent burning velocity, m/s
T temperature, K
V volume, m3
 xvi

Rred dimensionless
maximumpressure(Ped/Pj)
ill, dimensionless
vent burstingpressure(P,/Pi)

Greeksymbols

turbulent enhancementfactor

7 ratio of specificheatat constantpressureandvolume 1-4)

PU unbumtgasdensity
Pb burntgasdensity
X/P deflagration-outflow-interaction
number
& expansibilityfactor
(D equivalenceratio
AP pressuredifferential
 xvii

ACKNOWLEDGEMENTS

Firstly, I would like to thank and extendmy heartfelt gratitudeto my supervisor,Prof.

G.E.Andrews for giving me the opportunity to undertake this study. His constant
involvement,advice, brilliant thoughtsand encouragements throughout this project is
gratefully appreciatedandacknowledged.

Thank is also expressedto R.A. Boreham, for his help and technical expertise for
completingthis project.I would like to thankmy other supervisor,Dr.Roth Phylaktou,Ms
S.K Willacy for her help and motivation and fellow postgraduatesfor guidance
throughoutthis project,

Most especiallya special thank to my parents and family for a constantsupport and
encouragementin completion of this project and thesis. Not forgetful to Nadiah, Liz,
Toby, Emma,Danial,Brian and Bev for welcomedistractionfrom study and making my
life in Leedsmoremeaningful.

And to God,who madeall thingspossible.

 xviii

PUBLICATION BY THE AUTHOR

Kasmani,R. M, Willacy, S. K, Phylaktou,H. N and Andrews, G. E. (2006), Self

acceleratinggasflamesin largeventedexplosionsthat are not accountedfor in current
vent design, 2nd International Conference on Safety &
Environmentin ProcessIndustry,Naples,Italy.

Kasmani, R.M, Andrews, G.E, Phylaktou,H.N, Willacy, S.K. (2007), Vented gas
explosionin a cylindricalvesselwith a vent duct,Third EuropeanCombustionMeeting
(ECM 2007),Crete,Chania.

Kasmani,R.M, Andrews,G.E., Phylaktou,H.N. and Willacy, S.K.(2007), Influenceof

static burst pressure and ignition position on duct-vented gas explosions, 5'h
InternationalSeminaron Fire andExplosionHazards,Edinburgh,UK.

Willacy, S.K., Phylaktou,H.N., AndrewsG.E., Kasmani,R.M. and Ferrara,G. (2006),

Stratified Propane-Air Explosions of Global Concentration Outside Normal
Flammability Limits, 2nd International Conference on Safety &
Environmentin ProcessIndustry,Naples,Italy.
 CHAPTER I

VENTED GAS EXPLOSION - THE RESEARCH

 I

1.0 Introduction

The potential explosion hazard resulting from deflagration of gases in processing

systemis a reality which mustbe recognisedand consideredwhendesigningplantsand
processes.When the potential explosion hazardis not properly addressed,the result
would be catastrophic;both in termsof risk of injury to personneland propertyloss as
well asequipmentdamage.

Severaltechniqueshavebeendevelopedto preventthe destructivedamageto plants in

industries.Oneof the most recognizedand widely usedexplosionprotection strategyis
venting.Currentgas explosionvent designstandardsin the USA (NFPA68,2002) and
European(2007) rely on the vent correlation first published by Bartknecht in 1993
(Siwek, 1996).NFPA 68 alsorecommendsthe correlationof Swift (Swift, 1983)at low
The correlationsapply to compactvessels,which Bartknecht definesas
overpressures.
those with a length to diameter ratio, LJD < 2. The European Directive on 'Gas
Explosion Venting ProtectiveSystems'(2007) becamemandatoryafter August 2007.
The limit of LID of 2 given in Bartknecht'sequationas the upper limit for a compact
vesselin order for the sphericalflame propagationis still valid assumptionfor most of
the flametravel.

Studieson vented gas and dust explosionshave shown the evolution of venting area
with pressuredepend on the nature and state of the initial explosive mixture i. e.
composition,initial pressureandtemperature,pre-ignition turbulenceand on the vessel
characteristicsi. e. dimensionand shape,position of the ignition source,location, size,
strengthand shapeof vent, presenceof obstacleswithin the vessel.An understandingof
the mechanismsby which pressureis generatedin ventedexplosionsis importantin the
design of explosion reliefs and to the investigation of incidents. Such knowledge
providesthe basicof the developmentof prediction methodsfor use in risk and hazard
studieson an actualor proposedindustrialplant.

The vented gaseousdeflagrationshave been studied extensively for many years to

provide understandingof the phenomena.However the information on explosion
venting is mostly empirical and relatedto experimentsconductedwith homogeneous,
 2

quiescent gas-air mixtures. Yet, the different basic physical aspects of vented
deflagrationshave still not beenstudiedsufficiently. For decades,venting experiments

witb various initial conditions and modelling model using lumped parameteras a
venting parameter were done to understand the dynamics and physical process of
venting. From the intensive works done by previous researchers(Bartknecht, 1993,
Bradley andMitcheson, 1978a,Bradley and Mitcheson, 1978b,Burgoyne and Wilson,
1960, Cousins and Cottons, 1951, Cubbage and Marshall, 1974, Cubbage and
Simmonds, 1955, Cubbage and Simmonds, 1957, Donat, 1977, Zalosh, 1980),
correlations were offered to be used in order to design the required vent area;
unfortunatelythey are strictly applied within the limits of the experimentalrange
investigated.Hence,the key knowledgeaboutthe physicaland dynamic mechanismof
venting with or without the relief pipe needsto be understoodin order to reduce the
potential of explosion severity.

1.1 Venting gas explosions

1.1.1 Generaloverview

Early work on explosionrelief wasdoneby Cousinsand Cottons(Cousinsand Cottons,

1951)which usedpropaneandhydrogenin four vesselshaving a rangeOfLID ratios of
1.41to 22.1.They studiedthe effect of ventburst pressuresand vent ratio on maximum
reducedpressureinside the vesselat different initial conditions.The work concluded
that initial condition and strength of vent burst pressure caused the increase of
maximumpressureinside the vesselduring venting and gaverecommendationfor vent
area design. Another early study was carried out at the Midlands ResearchStation
(MRS) of the GasCouncil by Cubbageand Simmonds(Cubbage
and Simmonds,1955,
Cubbageand Simmonds, 1957). Their thorough work concernedwith the explosion
venting of industrial drying ovens using town gas as the fuel. The experimentswere
done on oven which were approximatelycubical and with variety of vent condition
coupling the influence of vent cover inertia. The significant of their work resulted in
two peaksof pressuresfrom the explosionwhich later beenused to correlatethe vent
arearequirementwith the reducedpressureobtained.The magnitudeof eachpressure
peak has also beenquantified in terms of vent ratio relief panel failure pressure,mass
 3

per unit areaof the relief, location of ignition as well as shelf (obstacles).
arrangement
Low breakingburst pressureslead to large flame elongation,and hencelarger name
areaand result to higher expectedvesselpressureswhereasat high breaking pressure
the effectof flamedistortionis reduced.

Otherexperimentalstudieswhich haveyieldeddatasetsagainstthecorrelationsthat can

be testedincludethoseof Harris (Harris, 1967),Harris and Briscoe(Harris and Briscoe,
1967),Pritchardet al (Pritchard,Allsopp and Eaton, 1995)and Zalosh (Zalosh, 1980).
From theseworks, it has led to the developmentof a numberof empirical and semi-
empirical methodsand of the scalinglaws. The correlationmethodshave beenwidely
used since it is obviously desirable,for design purposes,to be able to accessthe
effectivenessof the explosion relief in a given situation by calculation rather than
experiment.

The introductionof gasnomogramsbasedon Bartknecht(Bartknecth,1981) has been

recognisedto be most commonly usedin predictingthe requiredvent areaas they are
simpleto useandcan be extrapolatedto the maximumlimit, i. e. volume up to 1000M3.
However, the nomograms do not compare well with actual experimental data,
introducingseriouserrorsasthe graphsbeingdifficult to readaccuratelyandgave over-
predicting valuesfor interpolation and extrapolationoperation.Besides,they can only
be used with confidencefor four typical fuel-air mixtures, over a narrow range of
burstingpressuresas being investigatedby Simpson(Simpson,1986)while only semi-
empirical recommendationsare given for their extrapolationto difkrent conditions.
Simpsonalso noted that the low pressureextrapolationof Bartknecht's nomograms
would only conservatively yield high relief area estimates.Since the nomograms
introducedmoreerrors,equationsaremorefavouredandnomogramstake secondplace.

The most usedand recommendedcorrelationfor venting gasdeflagrationis adoptedby

NFPA 68 which basedon the works of Swift (Swift, 1983)and Bartknecht(Bartknecht,
1993).Bartknecht's correlation is given as the vent area,A, as a function the vessel
volume, V213multiplied by a complex term that includes the mixture reactivity KG,
reducedoverpressurein the vented explosion,Predand static vent burst pressure,P,
The ratio of V213/A,has been referred to in the explosion venting literature (Harris,
1983)asthe vent coefficient, K. The V2/3dependence
of overpressureon the test vessel
 4

volume is a characteristicof sphericalor compactvesselexplosions,where the flame

remains mostly in spherical shapeduring venting processand this is the mandatory limit
for L/D =2 in European Standard (2007) and NFPA 68 (NFPA68,2002), If the

sphericalflame propagatesat a constantrate irrespectiveof the vesselvolume, there

Of Predon volumeotherthan K. However,Bartknecht's
shouldbe no other dependence
equationrely only on his set of experimentaldata at different volumesand vent area
(Bartknecht, 1993). At a fixed gas or mixture reactivity and fixed volume, Bartknecht

correlated the experimental results as shown in Eq. 1.1 where a and b given in Table 1.1.

-b (1.1)
A, = aP,,
d

Table 1.1Valueof a andb for Eq. 1.1

Gas KG b
a
Methane 55 0.164 0.5720
Propane 100 0.200 0.5797
Town gas 140 0.212 0.5900
Hydrogen 550 0.290 0.5850

Whenthe a valuesbeenplotted asa function of the gasmixture reactivity, KG,it gave

A, = (0.1265 log KG - 0.0567)Pd -0.5817 v 213

(1.2)

Kr,= (dPIdt)V" bar-m/s (1.3)

where (dP/dt)is maximumrate of pressurerise andV is (m3). It shouldbe noted

volume
that the constantsused in Eq.1.2 were determinedfor V= 10 rn3 and the value of Kcj
givenin Table 1.1werenot taking into accounton the volumeeffect nor the valueof b.

The correlationdevelopedby Bartknecht(Bartknecht,1993)can only be applied on its

own limitation asthe equationis the empiricalderivedfrom his experimentalworks and
no other publishedventing data included as notedby Siwek (Siwek, 1996)in his study
on publishedcorrelationsin ventedgas and dust/air-mixtures. In order to validate the
equationdeveloped,other data should be included so that the equation can be used
globally with confidence.
 5

NFPA 68 adopted Swifts equation (Swift, 1983) to apply for design and scale-up of
it in
venting system purposes and seem that the results obtained were good agreements
data obtained by Donat (Donat, 1977) and Harris and Briscoe (Harris
with experimental
Briscoe, 1967). Again, there is a limitation in using this approach as it can be
and
i. Pred is not more than 200 mbar and only suits to low-
applied on certain conditions e.
enclosures. Meanwhile, Swift's correlation is given as A, as a function of
strength
cross-sectionalarea of the A,
enclosure, multiplied to reactivity terms which include the

burning velocity, S,, and turbulent enhancementfactor, P. The use of the vented vessel
in
surface area vent correlations was first introduced by Runes (Runes, 1972) as the

maximum possibleflame to
area and was used calculatethe maximum flame mass
buming rate.

In Bartknecht'scorrelation (Bartknecht, 1993) the gas/airreactivity is given by the

mixture reactivity index, Ko and burning velocity, S,, by Swift's correlation (Swift,
1983)which also usedin most modelsof turbulentexplosionsand previous ventedgas
explosioncorrelations(Zalosh, 1980).The Kr, is not constant for a given gasbut varies
with composition,temperature,pressuresand vesselsize. It is easily shown that KC;is
directly proportionalto the burningvelocity, S,,in the latter stagesof a sphericalclosed
vesselexplosionas correlated by Kumar (Kumar, Bowles and Mintz, 1992) and Nagy

(Nagy andVerakis, 1983) and hencethe two parametersto quantify the reactivity are
directly related.

Some confusionhas beencausedby the fact that correlationformulae resulting from

these two different approaches,Su and Kr, in Swift and Bartknecht's correlations
respectively,imply different results to predict the vent area. More specifically, the
introduction of the venting deflagration index, Y.0 has been seen by some as
counterintuitive and therefore,judged incorrectly. The dependent
of Ko on volume
appeared to be a reflection of the tendencyof the flame ball to increaseits effective
surfaceareathroughinstabilitiesin the flame sheetstructure;the flame sheetgeometry
departsfrom an initially smoothsphericalform to one of the greatereffective surface
area acrosswhich combustioncontinuesat S,,which prevails at the local conditions of
temperatureand pressure(Pritchard,Allsopp and Eaton, 1995).The sameobservation
 6

of volume dependence
on KG was also madeby Chippet (Chippett,1984)on his work,
on propane/airmixtures.

Meanwhile,the use of S,,as the fuel characteristicis preferredas it is independentof

equipmentgeometrybut is moredifficult to measurethan KG,Cubbageand Simmonds
(Cubbageand Simmonds,1955,Cubbageand Simmonds,1957),Cubbageand Marshall
(Cubbageand Marshall, 1974) and Bradley and Mitcheson (Bradley and Mitcheson,
1978a, Bradley and Mitcheson, 1978b) used S,, as the basis of the their venting
correlationin orderto establishthe relationshipbetweenthe pressuredevelopedin the
combustionphenomenaduring venting. Theoretically, the flow pressureof unburnt
gasesthroughthe vent is proportionalto týhesquareof the productof burning velocity
and vent ratio. Thus,the higherthe burning velocity of the gasinvolved, the higher will
be the maximumpressuregeneratedin the enclosure(i.e. S,,for hydrogenis 3.1 M/s and
S. for propane is 0.45m/s). However, the fundamental burning velocity refers

significantly to the maximum rate of propagation of a flame into a quiescent gas

mixture of the appropriate(near-stoichiometric)concentration,and initially at ambient
conditions of temperatureand pressure.In determining the burning velocity, some
works had scatteredvalues for high reactivity gases i.e. ethylene and hydrogen
(Andrews and Bradley, 1972).This theoreticalideal situationdid not encounterin gas
explosion;turbulentis generatedas the unburntgas is pushedaheadof the flame front
when the vent is openedwhich can give rise to the additionalback pressureor driven
past the obstaclesor over surfaces.The effect of turbulenceis well known and been
investigatedby various past researchers(Molkov, Dobashi, Suzuki and Hirano, 2000,
Munday, 1963,Yao, 1974).Thus,becauseof this fundamentalproblem,the useof KGis
suggestedby NFPA 68 in order to determinethe required vent area for protecting the
And yet, sincelessdataof Kr, are published,only data given
processvessel/enclosures.
by Bartknechtwork (Bartknecht,1993)and NFPA 68 (NFPA68,2002) recommended
be usedasa reference.

Other correlationshave beendevelopedby severalresearchersin order to improve the

design requirementand agreewith experimentalwork i. e. Runes,Rust
and Simpson
Rasbashand Cubbageand Marshall (Lees, 1996) but, it
was either fragmentary of
obtained under different conditions. Since all correlations only applied to certain
condition and with a limit of validity, Bradley and Mitcheson(Bradley and Mitcheson,
 7

1978a, Bradley and Mitcheson, 1978b) have compared the results obtained by lumped-

parameter model of vented explosions to the experimental data but yet, only a
quantitative agreement was observed. The work was regarded more complete since it
also offers a sound theoretical basic in predicting vent area, however the criteria they
developed have been shown inadequate for large volumes with obstacle-generated

turbulence (Swift, 1983). Further critical examinations of various existing equations and

especially of their extrapolation within and beyond their recommended validity range
have been presentedby Molkov el al (Molkov, Dobashi, Suzuki and Hirano, 2000) who

also developed a new so-called 'universal correlation' for vent area calculation. Razus
et al (Razus and Krause, 2001) reviewed the empirical correlations developed for vent
sizing based on the reduced pressure parameter. Following the authors, equation by
Bradley and Mitcheson (Bradley and Mitcheson, 1978a, Bradley and Mitcheson, 1978b)

and Molkov (Molkov, Korolchenko and Alexandrov, 1997) gave relatively close
predictions to experimental data for most common hydrocarbon-air mixture.

Since published correlations are used to predict the overpressuredeveloped in

deflagration vented vessel with a validity restricted to certain limits of, parameter
characteristicfor the vessel,the vent and the explosivemixture, there are no generally
acceptedcorrelationsto guide the determinationof vent areasfor gaseousas being
establishedfor dust/airsystemi. e. VDI 3673.

Nevertheless,relevantexperimentaldataaremostly limited to small aspect-ratiovessels

usuallycubical in shape(Cooper,FairweatherandTite, 1986,Cubbageand Simmonds,
1955,Cubbageand Simmonds,1957,McCann, Thomas and Edwards, 1985, Thorne,
Rogowskiand Field, 1983) and very little on other geometries.They both studied the
physicaland gasdynamicsprocessin ventedexplosionin order to discussthe effecting
factorsresult the numberof pressurepeaksobtainedinside the enclosures-,
however,the
effect of vesselandvolumeshapewerestill not be accounted.

Further, a special deficiency of experimentalobservationsexists for elongatedlarge-

aspect-ratiovessels, which are typical for industrial applications. The large scale
methane-airexplosion test performedat Rusfoss,Norway by Moen et al (Moen, Lee,
Hjertager,FuhreandEckhoff, 1982).The test found that obstacleseven relatively small,
havea dramaticinfluenceon the violenceof the explosionand pointed out that initially
 8

quiescentexplosivemixture criterion was totally for

inadequate large scaleexplosionin

obstacleenvironment. Flame propagationin such vesselshas beenproved to be very

differentfrom thosewith small-aspect-ratio
sphericalvessels.

Yet, therefore,some confusionconcerningthe applicationof the various equationsto

practicalsituations,both as to which is the most appropriateequationto use in a given
circumstancesandalso asto the rangeof valuesof the variousparametersfor which the
equationare valid. It is worth noting that the disparity betweenthe experimentalvents
area- rangingfrom small to larger-closedto-room enclosuresand correlatedvaluesare
due to the uncertainty on the fundamentalbasis of vented explosion and lack of
publishedexperimentaldatathat would renderthe developmentof vent sizing design.

Modelling on venting explosion has been carried and still, disparity betweenmodels
proposedis huge. Complex computer codes requires considerableinput data and
accuratethermophysicalpropertiesas well asaccessto super-computerfacilities are of
little use in the systemdesign (Zalosh, 1995).Their utility in exploring less complex
flame laws however,is not in question.Designtechniquesshouldbe quick and easyto
use,accurate,conservativeyet realistic andprovideresultsthat accordwith experience.

1.2 Overview of the project studies

Since the publishedguideline of NFPA 68 (NFPA68,2002) and EuropeanStandard

(2007) has only included Bartknecht'sequationfor L/D =2 which only his own data
was used in developing the correlation, there is no guaranteedthat empirical
correlationscould provide practical and safe venting design for' industrial purposes.
Further, the variation of fuel/air mixtures concentrationis not widely studied, where

manyexperimentaldonein stoichiometricconcentration

The publishedgas venting data from 1920sto date were complied and collected with
different volumes, gas reactivity, ignition position, shapes and different initial
conditions.From the reporteddata, the comparisonbetweenBartknecht's and Swift's
equation were made as those equationswere offered in NFPA 2002 and European
Standard(2007). Further, those data were analysedto be comparedto Bradley and
 9

Mitcheson'sequation(Bradley and Mitcheson,1978a,Bradley and Mitcheson, 1978b)

as well as Molkov's equation (Molkov, 2001). Theoretically, for a vent to give no
increasein overpressureother than due to the pressuredifferencecreatedby the mass
flow of unbumtgasthrough the vent, the vent massflow rate is consideredto have an
equal to the maximum mass burning rate of the flame. In the simplest case, the
incompressibleflow equationcan be used for low Predwhile treating the vent as an
orifice with dischargecoefficient, Cd which dependson the designof the vent and K.
Basedon this approaches,two different methodswere proposednamedas Method 1
andMethod 2. Method I assumesthat the maximumoverpressure occurswhenthe vent
unbumtgas flow is is
rate at a maximum and equal to the maximum massbum rate at
the flame front whereas Method 2 based on the assumption of the maximum
overpressure occurswhen the vent unburnt gas flow rate is at a maximum and this is
equalto the maximumunbumtgasdisplacedflow by the flame front.

The differenceof these two methods was the use of S,, and S,, (E-1) terms in the
equations.The simple approacheswere evaluatedwith additional of compressibility
factor and turbulencefactor to simulatethe real explosions.Thesetwo methodswere
comparedwith the publishedexperimentaldatato justify the applicability, validity and
limitation with respectto the practicalapplication.The detail discussionon the effect of

volumeand shapeof the vessels,turbulent enhancement factor, P and K were madein

orderto investigatethe importantfactorsinfluencingthe Pr'din the developmentof gas
ventingexplosions.Further,to demonstratethe applicability of the proposedmethodsto
the practical application, seriesof vented gas explosionsin two cylindrical volumes
werecarriedout, mainly for mediumandsmall scalevolumesi.e. 0.2 and0.0065m3.

1.2.1 Simply venting gas explosions

By performing the experimentswith simply venting, the physics and dynamics of

explosive mechanismwere investigated. The project focused on the two different
volumes of cylindrical vessel ; medium and small size volume to highlight the
occurrenceof self-accelerationeffect in mediumvesselcomparedto small vesselwhich
is not takeninto accountin ventingdesignguide.
 10

Intensiveworks in venting gas explosionhad been doneby Cubbageand Simmonds

(Cubbageand Simmonds,1955, Cubbageand Simmonds,1957) on industrial drying
ovenfor varietyof fuel gasesandflammablevapoursin enclosuresup to 14 M3volume.
They reportedthat there are two distinctive peakpressuresobtainedfrom the pressure-
time profile explosion development.Hence, from this results, the correlation which
containstermsexpressingthe effect of fuel/air mixturescharacteristicsand the volume
enclosuresin which the explosionoccurs.This equationcan be applied to any fuel/air
mixtures sincethe effect of the combustioncharacteristicsof different gaseson the
pressuregeneratedis allowedfor by the valueof burningvelocity, S,,,

Bartknecht's (Bartknecht, 1993) work on venting has been adapted in NFPA 68

guidelinesfor ventingdesign.His works involved vesselvolumeup to 60 M3vesseland
usedonly four different fuels which were methane,propane,hydrogenand town's gas.
Theseexperimentsthat been evaluatedwere carriedout with restrictedconditions and
the equationcanonly be appliedif the requirementsfit. Someof the conditionsneeded
to be fitted into the equationare P, is between0.1 to 0.5 barg and LJD of 2 which the
currentprojectclosely followed. It is the aim of the currentproject to fit the condition
requiredin order to investigatethe parametersthat not beentaken into accountin the
equationdeveloped.In a ventedexplosionwithout a vent pipe, the peak overpressureis
relatedto two main pressurepeaks,the first is relatedto the vent static burst pressure
andthe secondto the pressureloss of unburntgasthroughthe vent.The secondpressure
peak increasesas the vent coefficient, K (= V2/3/A,), increasesand the first pressure
peakdominateswhen the static burst pressureis high. In the presentwork a relatively
high K of 16.4(I/K = 0.061) was usedand the overpressurewithout the vent pipe was
dominatedby the pressureloss of unburnt gas through the vent. The Bartknecht's
correlation(Bartknecht,1993) for vent area is in terms of I/K and for a static bust
pressureof 100mbthe correlationtakesthe form of Eq. 1.2.

1/K [0.1265logK() - 0.05671/Predo'5817 (1.4)

=

KG is the reactivity parametergiven by the maximum rate of pressurerise in a 5-litres

sphericalclosedvesselmultiplied by the cuberoot of the volume.Bartknechtalso gives
the venting correlation for propane,methane,hydrogen and town's gas/air with 100
mbarvent burstpressurerespectivelyas:
 11

0.5820
1/K, = 0.197/Pd (1.5)
0.5820
1/Kv = 0.164/Pred (1.6)
0.5820
VK, = 0.291APýd
0.5820
1/Kv = 0.220/Pd

It will be shown that these predictions are grossly in error and even with a vent pipe the

overpressures come no where near these values. When the experimental data of
Bartknecht is examined (Kasmani, Willacy, Phylaktou and Andrews, 2006, NFPA68,
2002), it is clear that the exponent Of Predis fitted to the data with I/K > 0.2 or K<5

and the be
correlationsshouldnever usedfor higher K. This is not recognizedin NFPA
68 or in theEuropeanguidanceon venting.The limitation on the correlationis statedas
Pred<2 bar andthe experimentaldatashowsthat this is effectivelya limit of K of 10. It
is found that in cubic vessels,this condition is not valid as the flame have touchedthe
wall well before2 bar overpressurewas reached.However,none of the experimental
data for Pred> 0.8 bar fits the correlation and all is lower than the correlation, This
effectively givesthe validity of the correlationat K<5 andPred< 0.8 bar andthis is the
incompressiblevent flow regime. When the nearestBartknecht's data point to the

presentresultsis used,which is aI M3 vesselwith I/K = 0.058 and the overpressure

with a 100 mbar vent burst pressurewas 2.5 bar for propane/air.It is clear in the
Bartknecht'sexperimentaldata,there is a volume effect additionalto that in K that is
not includedin the correlation(Kasmani,Willacy, PhylaktouandAndrews,2006) andis
likely associatedwith flame self-accelerationdue to the developmentof cellular flames.
This effect wasalsoimportantin the presentwork.

In this work, the effect of fuel/air mixtures concentrationat different ignition position

will be investigatedas little attention is given in this aspect.As stated in ATEX

Directive (94/9/EC) venting equipment should be protectedbased on worst case
,
scenarioand it is found that stoichiometric concentrationat centrally ignited is not
alwaysthe worstcase.

Further, the use of vent cover i.e. lightweight relief panel, rupture membrane or
explosiondoor has proven to have effect on explosion pressuregeneration inside the
i. e. back pressure due to inertia of the vent cover (Molkov,
vessels/enclosures
 12

Grigorash, Eber, Tamanini and Dobashi, 2004). This commercial vent cover can be

affected by the inertia of construction elements moving by overpressure during the

explosion. The example of commercial vent cover available for venting is shown in

Fig. 1.1.

Figure 1.1 Example of commercial vent covers and its application.

The effect of inertia has been studied extensively in order to quantify the effect of
inertia to the Predduring venting (Bartknecht, 1985, Cubbage and Simmonds, 1955,
Molkov, Gfigorash, Eber and Makarov, 2004). However, in this present work, the effect

of vent cover inertia can be neglected as the material used is easily broken and not
impending the flame propagation during venting explosion.

1.2.1.1 Self-acceleration of spherical flames

Although the laminar burning velocity and hence the flame speed in a spherical flame at

constant pressure is normally assumed to be a fundamental constant a particular gas/air

mixtures. However, experiments in vessel of different size has shown that this is not the
case (Chippett, 1984, Pritchard, Allsopp and Eaton, 1995). The spherical flame speed
has been demonstrated to vary with the flame radius, accelerating when the flame is

small then having a period of near constant flame speed and at so-called ' critical flame
radius', the flame starts to rapidly accelerate through a mechanism called ' self-
acceleration' or flame cellularity. The net result is the KG is a function of the vessel size.
Harris and Wickens (Harris and Wickens, 1989) had done series of small and large scale
 13

balloon tests on natural gas/airmixtures.They have shownconclusively that the high

flame speedsof > 100 m/s are impossible to achieve in unobstructedexplosion
comparedwhen the flame propagatesthrough the unconfined explosion. In small

volume balloontest, the flame speedof naturalgasis found to be 3.1 m/s,closeenough

to the valuegiven by Andrews andBradley (Andrews and Bradley, 1973)of 2.6 m/s for

a 50 mm diameter sphericalflame in a 300 mm diameter closed cylindrical vessel.

However,the increasevalueof flame speedin 3 rn diameterof a large balloonin respect
to its laminarburningvelocity wasobservedin naturalgasandethyleneLe.7 and 15 M/s
respectively. From it be
the results, can said that all flames have a minimum distance

that they propagateat the laminar flow rate before startingto self-acceleratedue to the
processof flame instability throughthe formation of cellular flames that have a greater
total flame surfacearea than a smoothlaminar flame. The detail of self-acceleration
mechanism hasbeen discussedby Bradley et al (Bradley,Cresswell
and Puttock, 2001).
In cellular flame, the magnitudeof the increasein flame speeddependson the type of
gasand the equivalenceratio of the mixtures.It is generallyknown that methanehas a
smaller accelerationthan propanewhich in turn, is smaller comparedto hydrogen.
Hydrogen/air flame becomescellular very easily as been shown by Andrews and
Bradley (Andrewsand Bradley, 1973)in 50 mm diameterof closedcylindrical vessel.
They observedthat the onsetof flame cellularity occursat the mixtures leanerthan 25
% concentration.

Chippet (Chippett, 1984)introduceda methodfor the inclusion of the developmentof

cellular flame in designprocedures.In his correlation, 77is consideredas a factor for the
increasein flame surfacearea due to cell formation. This is directly proportional to
Prandtl number,Pr and Reynoldsnumber,Re. Pe.Re is the flame stability parameter
given by Istratov and Librovich (Istratov and Librovich, 1969) which predicts the
critical conditionsof the onsetof flame instability and cell formation. It also postulated
factor, P (= ST/Su).Thus, it
that q is essentiallyequivalentto the turbulentenhancement
is important to understandthe flame self-accelerationand to include the parameterin

explosionprotectiondesign.
 14

1.2.1.2 Implication of the objectivesto the presentstudy

Fromthe discussionabove,the studywas performedto gain betterunderstandingof the

physicalprocesseswhich are important in the generationof the and
overpressures to

elucidatehow this processesinfluence the severity of an explosion. Simplification on

design is still preferred for venting design but it needs to be theoretical and
vent
practical basis which can be explained through the actual experimental results.
Generally, this project's objectives are;

* To study the influence of K parameteron Predin terms of vessels' shapeand

the
volumeof the enclosuresand effect of burstingpressure,P,
-
To quantify the influence of turbulent enhancementfactor, P and flame self-
acceleration on vented gas explosion for different gas reactivity and
concentrations.
To proposea simpleapproachin calculatingthe Predand vent area,Av using two
different methodswhich crucial parametersinvolved in venting gas explosion
developmentis included.

1.2.2 Vented gas explosionwith relief pipe attached.

The studyof the venting gasexplosionof different fuel/air mixtures and concentrations

were further examinedin terms of the effect of duct pipe attachedto the main vesselon
P ,,. Basedon the experimentalanalysisof ventingexplosionwith and without a pipe,
it is known that the severityof the explosionis likely to be 2 or 12 fold increasewith the
presenceof a duct to
with respect simply vented vessels (Bartknecht, 1981, Ferrara,
Benedetto,Salzano,Russo,2006, Ponizy and Leyer, 1999a,1999b,Siwek, 1996). The
resultsof DeGood et al (DeGood and Chartrathi, 1991) and McCann et al (McCann,
Thomas and Edwards, 1985) for various duct lengths supportedthese results. The
surprisingfeatureof theseresultsis that the length of the duct is not a major parameter
andonly short and long ductswere separatedby Bartknecht(Bartknecht,1993).This is
becausethe increasein the overpressureis dominatedby duct entry and exit pressure
losses;a duct would have to be about 100pipe diameterslong before the flow friction
losseswereequalto the combinedinlet and exit pressurelosses.The presentwork used
 15

relatively a relatively short duct of lm long and 0.162 m diameter (UD = 2) ut it

.
exhibitedall the featuresof explosionventingwith a vent duct In
attached. Bartknecht's
(Bartknecht, 1993) 3
the in long pipe was 0.5m diameter and hence the LJD was 6,
work
in As friction lossesare a function of pipe LJD
closeto that used the presentwork. pipe
a dependence of the vent overpressure on the L/D would be expectedratherthan simply
on the lengthof the duct, irrespectiveof its Particular
diameter. emphasiswasplacedon
the determination
of flame speedsupstreamof the duct, as well as flame speedsand gas
in
velocities the vent duct.

Further,the problemseemsto be more complicatedif the vent coverswere placedat the

vent, The effect of bursting vent on pressuredevelopment'in vented explosion have
beenstudiedby severalworkers(Ferrara,Benedetto,SalzanoandRusso,2006,Pasman,
Groothuizenand Gooijer, 1974, Ponizy and Leyer, 1999a,Ponizy and Leyer, 1999b,
Cooper, Fairweather and Tite, 1986, Cubbage and Marshall, 1974), limit at
stoichiometricfuel It
concentration. is found that the peak pressureinside the vessel
during duct-ventedgas explosionsdid not result from the external explosion and not
occurwhenthe flame is propagatingin the duct asreportedby Kasmani et al (Kasmani,
Andrews,Phylaktouand Willacy, 2007a,Kasmani,Andrews, Phylaktou and Willacy,
2007b).Vent rupture also generatesa pressurewave which interacts with the flame
front to distort it and hence increaseits surfacearea and massburning rate. These
combinedeffectsof turbulenceandpressurewavescreated by the vent burstingresult in

accelerationof the flame prior to the vent duct and also increasethe flow velocity,
turbulence and flame speed in the vent duct. Tests with ducted explosion vents
generallydisplay Helmholtz oscillations, that is the pocket of burned gas within the
vesselundergoes bulk motion towards andaway from the vent opening,due to the mass
of the duct contentsandthe compressibilityof the gasin the primary enclosurethat acts
as a spring(McCann,Thomasand Edwards,1985).The occurrenceof a sharppressure
peak in the vent duct is also mentionedby Kordylewski and co-worker (Kordylewski

andWach, 1988)and maximumpressureeffect in the vesselwas found to occur with a

particular duct length, equal to about 12 diameters. It is known now that the
intensificationof the combustionin the vesselis driven by an impulse generatedduring
the bum-up in the initial part of the duct, shortly after the flame penetratesinto it
(PonizyandLeyer, 1999a,Ponizy andLeyer, 1999b,Ponizy andVeyssiere,2000)
 16

When relativelynarrow ducts are usedwith a sharpvessel-ductareachange,the flame

front enteringthe duct can be temporarily extinguisheddue to stretch (causedby the
inlet venacontractioneffect, which locally increasescentrelinevelocitiesby 64 %) and

cooling throughturbulentmixing with unburnedgas (lida, Kawaguchiand Sato, 1985,

Ponizy andLeyer, 1999a,Ponizy and Leyer, 1999b)which brings aboutstrongerburn-
up (i.e. with higher pressureamplitudes)during re-ignition (Ponizy and Veyssiere,
2000).

The increasein Predwith the additionof a vent duct is so largethat vent ductscannotbe
used without increasingthe vent area and duct size to achieve a reduction in the
overpressure.However,there is insufficient designdata for gasesto enablethis to be
done effectively and the physics of the processfor gas explosionventing is not well
understood.This contrastswith the situation for dust explosions,where a substantial
experimentaldatabaseexists(Lunn, CrowhurstandHey, 1988).The aim of the present
work was to explorethe physicsof the venting processwith a vent duct attachedwith
variousduct diametersto investigatethe effect of the vent staticburst pressureand the
duct diametersize for central and end ignition for an explosionvesselat the limit of
applicabilityof compactvesselventingcorrelationswith a lengthto diameterratio, LM
of 2. It is shownthat generally,reverseflow in ductsof largerdiameterswill result in
large scalemixing in the vesselwhich in turns generateshigherpressureincreasein the

vessel(Ponizy and Leyer, 1999a,Ponizy and Leyer, 1999b).It is consideredthat it is

the variation of the massbum rate and flame speedof the flame approachingthe vent
that has a strong influence on the vent flow and on the subsequentcombustion
behaviour.A major featureof the explosionsis that therearesubstantialproportionsof
the original flammablemixture in the test vesselafter the flamehasexited the vent duct.
This is larger for central ignition than for rear ignition. Kasmani et al (Kasmani,
Willacy, Phylaktou and Andrews, 2007) also shown that the faster mass bum rate
approachingthe vent as P, increasescausessonic flow in the vent and hencechoked
flow. This preventstherebeing any outflow from the duct until the pressurehasrisen in
the vesselto drive the bumt gasesout. In somecases,this condition leadsto a period of
mixed bumt gasand unbumt gas venting with micro explosionsand detonationsin the
vent duct. This phenomenahas been detailed by Ferrara et al (Ferrara, Willacy,
Phylaktou, Andrews, Benedetto and Mkpadi, 2005, Ferrara, Willacy, Phylaktou,
Andrews,BenedettoandSalzano,2005).
 17

1.2.2.1 The influence of duct pipe diameter

The influence of vent ducts on gas explosions was investigatedwith the aim of
determiningwhetherthe use of a vent duct of large areathan that of the vent, would
reducethe very large increases
in overpressurethat occur whenvent ducts are used.In

the presenceof a vent duct, an increaseof venting areaand duct diameterhas been
found to not alwaysresult in a decreasein the peak over pressure(Ponizy and Leyer,
1999a).

The main reasonfor the increasein the overpressurewhenlong vent ductsare attached
to vents is due to the phasein the explosionwhen the flame is in the vent pipe with
unburnt gas mixture aheadof it. The expansionof the burnt gases in the vent pipe
greatly acceleratesthe unbumt gas flow and this increasesthe vent pipe friction, inlet

and exit pressurelosses (Lunn, Crowhurstand Hey, 1988).These effects are a function

of the dynamic in
pressure the In
vent pipe. principle the dynamic in
pressures the vent
be by
pipe can reduced simply using a larger vent diameterthan that for the vent, rather
than increasingboth the vent and vent pipe sizes.For example,if the vent pipe was
twice the diameterof the vent then the vent pipe dynamicpressurewould be reducedby
a factor of 16, if the vent massflow rate remainedconstant.Some evidence that a larger

vent pipe diameter would reduce the overpressurewith no changein the vent size was
provided by Nagy (Nagy and Verakis), is
which quoted in NFPA 68 (2002). From the
experimentsperformedon dust explosionswith vent pipes (Lunn, Crowhurst and Hey,
1988),Hey (Hey, 1991) has suggestedthat the techniqueof using a larger vent duct
diameterthan the vent diameteris effective if the duct area/ventarea is about - 2-2.5
and when Predis less than 0.5barg. It is consideredthat this approachwould be a
simpler methodof designingfor safe vent pipes and the presentwork investigatedfor
gas explosionsa vent pipe that was close to twice the vent diameter,as recommended
by Hey (Hey, 1991).

1.2.2.2 Implication of the objectivesto the present study

From the discussionstatedabove,it is crucial to investigateand study the important

phenomenaoccurredduring the explosiondevelopmentin ventedgasexplosionwith the

 18

relief pipe attached to the primary vessel. It is the aim of the present study to quantify
the relationship of the Pnz,,,with ignition position, equivalence ratio and vent bursting

pressureP, as well as the duct pipe diameter and length in order to highlight the most
leading factors of the rapid increase in P.. at the presence of the duct in vented gas
explosion with respect to the simply vented explosion.
 CHAPTER 2

EXPERIMENTAL SET-UP AND

MEASUREMENT TECHNIQUES
 19

2.1 Test facility

The researchhas been carried out in Room BII which has been designatedas the
'Explosion Hazards-HighPressureTest Facility'. The site is shown schematicallyin
Fig. 2.1. Safetyconsiderationshavebeentop priorities to make sure that all tests were
carried out in a safety manner.Two separatedrooms have been designed: 'Control
room' and 'Test room'. Accessbetweenthe two roomswasvia an interlockeddoor that
formed part of the safety systemcontrolling the ignition circuit. The dump-vessel,test
vesselsand instrumentationequipmentswere locatedin the Test room. Data logging
hardwarewas situatedin this areaandwas electronicallylinked to a computernetwork
in the Control room. Data collection was synchronisedwith ignition using specialised
softwareknownasFAMOS.

2.2 Apparatus designconsiderations

2.2.1 Test vessels

In this projectresearch,threetest vesselswereused.Test vesselI was constructedfrom

cylindrical vesselwith internal diameterof 0.5 m and length of 1,0 m with vent areaof
0.021 m2 and the volume of 0.2 M3 For Test vessel2, the cylindrical vessel with
.
internal diameterof 0.126 m and lengthof 0.315 rn wasused.Further,for Test vessel3,
different length and diameterof pipes were connectedto the main vesselat constant
vent areaof 0.021 2
rn for ventedgas explosionwith relief pipe tests. Table 2.1 gives
detailsfor the test vesselsused.
 20

Test Room

Control room

Figure 2.1 Schematic picture of test facility and control room

 21

The previousworks (Bartknecth.

W, 1981,BradleyandMitcheson, 1978b,Bradley and
Mitcheson, 1978a, Chippett, 1984, Cubbage and Simmonds, 1955, Cubbage and
Simmonds,1957, Donat, 1977, Zalosh, 1980) have shown that the evolution of

pressurewith time in venting gas exýlosion dependson the nature stateof the initial
explosive mixture, i. e. composition, initial pressureand temperature,pre-ignition
turbulenceand on the vesselcharacteristicsi. e. dimensionand shape,position of the
ignition source,location, size,strengthand shapeof the vent, presenceof the obstacles
within the vessel.Further,the addition of the relief pipe to the vesselcan significantly
increasethe violenceof the explosionand the pressurein the vesselto the factor of 10
or morein which the explosionis venteddirectly to the atmosphere(Bartknecth,1981,
Kordylewski and Wach, 1988,Molkov, Baratov and Korolchenko, 1993, Ponizy and
Leyer, 1999a,Ponizy and Leyer, 1999b).The test configurationwas employedfor the
following reasons;

The presentproject was aimed at investigatingthe physical and gas dynamics

governingthe generationof pressurepeaks,Pred.Further,the influence

processes
of the ventcoverfitting at the vent on Predwill alsobe studied.
This configuration is complied with the UD restriction in NFPA 68 and
therefore,the results of this study will be comparedto the published venting
correlationsofferedby NFPA 68 andDraft of the EuropeanStandard.

2.2.2 Dump vessel

In order for safeventedexplosionsto be carriedout in the laboratory,the test vessels

wereconnectedto a large dump-vesselthat actedas an expansionchamber.Both burnt
and unbumtgas could then be safely ventedand continuedbefore vesselpurging. An
importantfeaturewas the fact that pre-testconditionsboth inside and outsidethe test
vesselscould be accuratelycontrolled.The overall dimensionsand thereforevolume of
the dump vesselwere also important variablesalong with the various vent diameters
andpipe fitting. This designprocesswas governedby the following factors;
The internal length and diameter of the dump vessel had to be such that its
volume was greaterthan that of the test vesselby the factor that allowed the
 22

explosions.If this was the case then the

simulation of open-to-atmosphere
presenceof the dump vesselwould have little or no effect on the explosions
development
in the test vessel.
9 Combustionof a gas mixture preparedin the test vessel would result in a
pressureincreasein the total systemvolume, VT that was madeup of the test
vesselvolume,Vt andthe dumpvessel,Vd suchthat,

VT = VI +Vd

Assumingadiabaticcombustionat a constantpressureof I atm,

Vb=
7.5V, (2.2)

where Vb is the burnt gas volume assuming an adiabaticexpansionfactor of 7.5.

Applying the ideal gaslaw,

V,
P2 = PI
V2
(2.3)

where

P2= absolute systempressureafter combustion (atm)

PI =I atm.

3)
Vi = systemvolumeafter combustion (Vb + Vd) (M

V2 = system volume before combustion (Vt + Vd) (M3)

Vd
P2 ='X 7.5V,+
(2.4)
V, +Vd

Hence, the total system overpressure,Psy,is given by;

P=6.5V, (2.5)
SYSV, + Vj
 23

This calculation was performed for each test vessel volume and values of Py, are listed
in Table 2.1.

2.3 Construction details

2.3.1 Test vessels

The Test vessel I was made up from the existing 05 in diameter and Im length
cylindrical test vesselsthat have been designedto BS1560and rated at 28 bars as
illustratedin Fig. 2.2. The vesselwas drilled with tappedbosseswelded on at positions
along their length,to allow instrumentationin the form of thermocouplesand pressure
transducersto be fitted. A removableblank plate was usedto close one end of the
vessels.This was centrallydrilled and tappedto allow a sparkplug, hoseconnectedto
the barometerandhoseconnectedto vacuumpump to be fitted. At the other end of the
vessel,a 0.162in diameterwas drilled to simulatethe vent areaof 0.012 m2.A certified
hydraulic pressuretest had been carried out on the test vesselsbefore commissioning
the tests.A scaleddrawingof test vesselis shownin Fig. 2.2 andmajor dimensionsand
pressurerating arelisted in Table2.1.

The Test vessel2 was madeup from 0.162 m diameterand 0.315in length cylindrical
test vesselsthat havebeen designedto BS1560 and rated at 35.5 bars (Fig. 2.3). The
vesselwas drilled with tappedbossesweldedon at positionsalongtheir length, to allow
instrumentationin the form of thermocouplesand pressuretransducersto be fitted. A
removableblank plate was used to close one end of the vessels.This was centrally
drilled and tappedto allow a sparkplug, hoseconnectedto the barometerand vacuum

pump to be fitted. Seriesof test on Test vessel 2 were carried out to investigatethe
effect of vent areaon explosiondevelopment.The single-holeorifice plate was usedto
simulatethe objectiveon varying the vent areawhich placedat the end of the vesseljust
beforethe gatevalve andit gives,

A, = (I - BR)xA
 24

where A, is the vent area, BR is the blockage ratio of the orifice plate and A is the area
of the plate.

For ventingwith relief pipes,Im pipe lengthhasbeenconnectedto the other end of the
Test vessel 1 for a constantvent areaof 0.012 m2 which beencalled a Test vessel 3
(Fig. 2.3). For Im pipe, thereweretwo different internal diameterpipe fitted, i.e. 0.162
m and 0.315 m. Vesselsupportwas in the form of moving framesfabricatedin-house
from L-sectionmild steel.Further,the test vesselwastied with the cradleof the craneto
easethe movementto facilitate quick replacementof vent cover andpipe length without
major rig dismantlement.

The openendednatureof the test vesselsnecessitates a methodof isolation during fuel-

air mixture preparation.This involved the closing of the gate valve (see Equipment
Specificationsection)that hasbeenbolted betweenthe flange at the open-endof the test
vesseland the flange of the dump vessel or the flange of the pipe if the pipe was
connectedto thetest vessel.A compressedair supplywith a line pressureof 10 bars was
providedfor valve operationwhich via solenoidvalve, could be manuallycontrolledby
a switch in the Controlroom.

Table2.1 Test vesseldesigndetails

Test vesselI Test vessel2 Test vessel3

Internaldiameter (nominal)(m) 0.5 0.162 0.5

Length(nominal)(m) 1.0 0.315 1.0

Wall thickness(m) 0.127 0.034 0.127

Designpressure(bar) 28 35.5 28

Flanges

Class(BS1560,1970) 300 300 300

Flangethickness(m) 0.068 0.0365 0.068

Numberof bolts 20 12 20
 25

Bolt-hole diameter (m) 0.042 0.022 0.042

Bolt-hole PCD (m) 0.670 0.2699 0.670

Diameter of bolts(m) 0.038 0.019 0.038

Pipe section

Internal diameter(m) 0.162/0.315

Section length (m)

Wall thickness(m) 3.4/3.4

Design pressure(m) 35.5/35.5

Length-to-diameter ratio(IJD) 6.17/3.17

-

Assembled test-vessel

Length-to-diameter ratio(UD) 2.0 1.94 2.0

Volume, Vt (m.3) 0.2 0.0065 0.22/0.28

Ratio to total system volume to

test volume, VT/Vt 260 8000 235/187

System overpressure due to

adiabatic combustion, P'Y' 24.9 0.812 27.5/34.6

(mbar)
 26

Li
--L-i

IIi

Test vessel I

N
i1
T1j
I.

/
Test vessel 2

Figure 2.2 Schematic drawing I'OrTest vessel I and 2

 27

cl

ý-- CO

Ln

(ID C)

,t
CU

LH

it
-`0

Figure 2.3 Scaled drawing of-Fest vessel 3 with pipc attached

 -)s

162
Z

Figure 2.3a S"iled drawing of'pipc section forTcst vessel 3.

 29

315

Figure 2.3b Scaleddrawing of pipe section for Test vessel3.

2.3.2 Dump vessel.

The durnp vesselwas designed in-11OLIse

as a Cylindrical sheli with dished ends. F'langcs

openings were included at various positiolls, with diameter matched that of the tost
 30

vessels.Their locations were determined by considering possible future research as well

s therequirementsof the presentproject.

Spacerestrictionslimited the major linear dimensionsof the dump vesselto 8 rn in

lengthand2.5 m in diameter.This gavea volumeof approximately40 m3. Single piece
torisphericaldishedendsof nominally 10mm thicknesswereweldedat both endsof the
vessel.The large 1500mm diameterflangedopeningsweresuppliedwith flat blanksto
close the vessel.Figure 2.4 shows scaled drawings of the dump-vesseland major
dimensionsarelisted in Table2.2.

Instrumentationports were fitted so that the explosionpressurescould be monitored at

variousdistancesfrom the test vesselexit. Fittings were tappedbosseswelded on the
dump-vessel back wall, opposite the test vessel connection points. Ports were also
located in the dump vesselfront-wall above the test vessel exit points.

Blanking plateswere drilled and tappedto allow connectionof pipe work to facilitate
evacuation via an external vacuum pump or the introduction of compressedair.
Pressureratedball valveswere usedto isolateconnectionancillary equipmentfrom the
dump vesselprior to ignition in the test vessel.Before explosiontesting was started,a
certified hydraulicpressuretest was carriedout by pressuringthe water filled vesselto
11.25 baro.

Table2.2 Dump vesseldesigndetails

Shell

Internaldiameter (m) 0.247

Length(m) 0.672
,
Shellthickness(m) 0.015

Torispherical(2:1) Dishedends

Outerdiameter(m) 2.500

Nominal platethickness(m) 0.010

 31

Assembled structure

Total length (in) 0.800

Design pressure(bar) 9

Certified pressure(hydraulic test) (baro) 11.25

Flanged openings

Type Nominal Neck thks Flange Number Bolt-hole Rating

bore bolts PCD (m)

NI 1.524 m 20 plate Special 52 1.759 Special

O/DIA

N2 0.508 m 10 plate RFSO 20 0,635 BS 4504

O/DIA 40/3

N3 0.162m SCH40 RFSO 12 0.2699 BS 1560

I class300

N4 0,0762m SCH40 RFSO 8 0.1683 BS 1560

1 1_ L class300

N5 1/4 " BSP ý COUPLING Special

 12

FIgLiFe2.4 DLIIIIJ)VCSSCI
SCI)CIIIItICCleSiglIS
 33

2.4 Vent cover selection

The vent coverswere selectedwith a rangeof vent burstingpressure,P,,., in between

0.1 to 0.5 barg.Four vent coverswere used;magazinepaper,aluminium foil, Melinex
membraneor Mylar film and I OOgA4 paper.The vent coverwasclampedbetweentwo
steelplatesbeforeit placedat the end of the test vessel.Using compressedair, the vent
burstingpressure,Pstatwas determinedexperimentally.The lists of the vent cover used
and its P, were listed in Table 2.3

Table2.3 List of P, of vent coversused

Material PV (expt) Thickness Average

(mm) mass/area

(kgIM2)

Magazinepaper 0.097 0.5 0.014

Aluminium foil 0.178 0.21 0.011

Melinex membrane 0.209 0.20 0.004

I OOgA4 paper 0.424 0.12 0.02

 LI

Figurc 2.5 VC11tCoVCI-, top left: MCIIIICX111C111bl",

', LISCdIrl the lests. I"I-0111 111C, 11 Lill]
1.1111111i
foil. From bottom left: I OOgA4 paper and magazine paper
L,

2.5 Instrumentation and data collection

2.5.1 Instrumentation techniques

Important parameters studied in the general analysis ol' gas explosions are tile I'lame

speeds and rnaxIIIILIIII PrCSSUI-C

attained during explosion development. Most

experimental Studies have included trallSdUcers 10 FrICaSUIT

CXI)IOSIOIIS
OVCI-p1-CSSUre.
Various types have been employed including those of' diaphragm and piezo-clectric

operation. For this present project, the relatively high pressure ocnerated and large scale
experimental rig meant that there were S111'ety 10RIICOut the ViSLIa11SatIOII
ColiSider,111011S
technique to monitor flarric travel.
 35

2.5.1.1 Explosionpressurehistory

As describedearlier,threadedbosseswereweldedalongthe test vesselswalls at drilled

positionsas locationsfor explosionsmonitoring equipment.The instrumentemployed
then becamean integralpart of the test apparatusand thereforewould be subjectto the
same internal pressure.The instrumentationtherefore had to be robust enough to
withstand the explosionpressureswithout losing its sensitivity. Absolute explosion
pressureswere measuredin the test vesselsand the dump vessel using Keller type-
PAA/I I piezoresistivepressuretransducers.
They hada5 bar measurement
rangewith a
rating of 10 bar. They had high sensitivity combined
maximumpressuremeasurement
with stability andshockresistance.Test vesselpressuretransducerswere locatedin the
ignition-endflangeandalongits lengthas shownin Fig. 2.2,2.3 and2.3b.

All the pressuretransducerswere calibratedusing a standardhydraulic dead-weight

calibrator.This was performedwith the transducersconnectedto the data acquisition
system.Calibrationof this integratedsystemeliminatedany errorsthat might arisefrom
the electroniccircuit connectingthe instrumentationto the datalogging system.

2.5.1.2 Measurementof flame positions

A thermocouplejunction placedin the path of a flame travel will register its arrival as a
changein voltagepotentialacrossthe junction. This is recordedas a distinct changein
the analogueoutput of the thermocouple.Accurate measurementof the time of this
signal changerelative to that measuredfrom an adjacentprobe allowed the average
flame speeds to be calculated at the midpoint between the two probes. This
thermocouple technique has been used to monitor the flame arrival in previous
explosion studies (Gardner, 1998) and the system was validated by comparison to
photographicrecords of explosions in a closed spherical vessel. Figure 2.6 shows
examplesof thermocouplesignalsas displayedon a computerscreenfrom which flame
arrival time measurementwere taken using a movable cursor, This thermocouple
techniquewas a primary methodof flame monitoring usedin this study.
 36

The thermocouple tips had an inherent thermal mass; however since all the
thermocoupleswere at the samespecification,it was assumedthat they had the same
responsetime. It shouldbe notedthat the thermocoupleswere not usedto measurethe
temperaturebut merelydetectingany changein temperature.The time delayor response
time, usuallyrefersto the time lag for full responseto the stimulusand therefore,is not
relevantat this point.

NtdxirrJaiG (1- 1.08

0.006

0.005

O.OCK

0.003

0.002

0.001

0
45

TIM(s)

Figure 2.6 Typical thermocoupletracesindicating the changein voltage output due to

flame arrival.

2.5.1.3 Thermocouples

Exposedjunction, mineral insulated type-K thermocouplesas illustrated in Fig. 2.7

which beensuppliedby FitmeeLtd. Oldhamwere used.The main body had a diameter
of 3 mm and was locatedthroughthe vesselwall so that the exposed0.6 mm-diameter
conductionwires were on the axial centrelineof the test vessel.Each test vesselwas
fitted with up to twelve thermocouplesalong its length. Threaded Swagelock

compression fittings were used to seal the units in tapped BSP bosses. The
thermocoupleswere locatedthrough the vesselwalls and so would be subject to high
dynamicloadsresulting from high velocity gas flow impact, before and after the flame
arrival. In order to limit the bending and maintain accurate positioning of the
 17

thermOCOUPIeS, structures were built into test vesselsas well as along the pipes
SUPPOI-t
that were used.

The support consisted of' aI rniri dianieter connecting wire, suspended above the axial
centreline of' the vessel f'rom the thin inctal strips at the end of' the vessel. The

percentage finear blockage of each strip was less than 0.4 I/r. CirCUlar Washers were

used to hold the thermocouples 15 rnm above their exposed junction.

! t-

Figure 2.7 Exposedjunction, mineral insulated type-K thermocouples Inside the rig.

2.5.1.4 Flame speed calculation

By dividing the distance between two thermocouples with the time difference in flarne

arrival, the averageflame ýpeedbetween the two thermOCOUples

COLlidhe calCUlatedand
assigned to the midway position. This method of flarne speed calculation was possible
from saved raw data from the thermocouples Upstream of the test vessels. From raw
data, all thermocoupies were normalised Usingcomputer command in FAMOS.

During the initial stage, flame speedswere relatively low but will dramatically increase

towards the downstream of' the vessels and to the connecting pipe. UsIng tile flame

speedsmethod described above, the flame speedswere calculated at some positions. In

order to extract precise results, a technique wits employed to eliminate the excessive
The rnethod involved 'sinoothing' the distance-time data by
flame speed 11LICtUations.

of' two ad.jacent points, The

successivecalculations of the flame speed at tile 111idpOint
smoothing processis illustrated in Table 2.4.

thbý
UNN ýI UbMýl
LEED6
 38

Table2.4 Procedurein calculatingthe flame speedsfrom raw data

Thermocouple distance-timesmoothing method

Raw data

Flamearrival time, T/C position,x Flamespeed Flameposition

top(ignition) XO'O

tj'O XI'O (xI'O-xO'O)/(tI'O-tO.

O) (x o)/2+ xo,o
j,o-xo,

t2,0 X2,0 (x2, o-x 1,0)/(t2, o-t 1.0) (X2,0-X LOY2 +X1,0

t3,0 (X3,0-X2,0At3,0-t2, (X3,0-X2,0)/2 + XZO

X3,0 O)

4.0 (X4.0-X3,0)/(t4,0-t3, O) (X4,0-X3, o)/2 + X3,0

X4,0

tn, O (Xn, O-Xn-1,0Xtn, O-tn-1,O) (Xn, O-Xn-i. o)/2 + Xn-1.0

Xn,O

Smoothing data

to,I= to'O X0,I= X0.0

tij = (to,o+tjo)/2 xij = (xo.o+ xj,o)/2 (XI., -x 0,1)/(tl,l - to,j) (x 1,1-x o,1)/2+ xo.I

(X2,1 t1j) (X2.1

t2.i= (t i,o +t2,o)/2 X2,i= (x 1,o + x2,o)/2 -X 1,1)/(t2,1 - -X 1,1)/2+ x1j

= 02,0 X3,1 = (X2,0 + (X3,1 t2j) (X3,1

t3j
+t3,o)/2 -X 2,1)/(t3,1 - -X 2.0/2 + x2j

X3,0)/2
tn, l " (tn-1,0 +tn, O)/2 Xn, I = (Xn-1,0+ (Xn, tn- (X + Xn-
l -X n-1,1)/(tn, l - n,I -X n-1,0/2

XnA/2 I'l) 1 1.1

As mentionedearlier,Fig. 2.6 shows flame position againsttime of flame arrival data

for 9.5 % CH4/airin a simply vented test for open venting in Test vessel I between
thermocouplesat the centre and at the exit of the vesselwhere the high flame speeds
were generated.The data series representsuccessive'smoothed' values using the
methoddescribedin Table 2.4. Thesethermocoupleto flame speedsproceduresneeded
 39

to get preciseand accurateflame arrival time as well as the flame speedsfor the tests
done.

2.5.2 Data acquisition

The short duration of the highly transient explosions meant that accurate analysis

required high speed data collection from the various instrumentations positioned
throughout the test geometry. Processing the large number of recorded data was aided
by specialisedanalysis software.

Initiation of datacapturedandthe time of ignition weresynchronisedusing Wind speed

Wavecapsoftwarewhich allowed parameterssuch as samplingfrequencyand pre- and
post-trigger samplingtimesto be varied.

Instrumentationwas wired to a 34-channelMicrolink 4000 systemwhich designedfor

capturinghigh speedwaveform with a sampling frequency of maximum 200 kHz per
channel.The systemmonitoredthe 34 analogueinputs comprisingpressuretransducers
Analogue/digitalconversionused2 bit ADC, giving a resolution of
and thermocouples.
I part in 212(= 4096). The voltage measurementrange for pressuretransducerwas
between0 to 100 mV. For a pressuremeasurementrange of 0-5 bar, the resultant
transducerresolution was ±1.2 bar. The stored digital data was transferred to the
computernetwork in the Control room. From this, subsequentsignal conditioning and
analysiswascarriedout usingFAMOS software.

2.6 Identirication hazards and safety measures

Hazardsidentification and handling were based on the experiencesgained over the

long-term, safe operation of similar casesand on advisesand input received from
University Health and Safety Officers. A number of possible hazard generation
mechanismswere identified and associatedsafety measureswere incorporatedinto the
designandoperation.
 40

2.6.1 Vesselfailure

The designfeaturesof the vesselin this study were listed in Table 2.1 and 2.2. All test
vesselswerepressuretestedandcertified to withstandpressureabove8 baro. This is the
maximum adiabatic overpressurethat can be generatedby any fuel-air explosion
initially at I bar. The possibletransitionto detonationcombustionwas also taken into
consideration.The actual design pressurefor each test vessel was higher than the
overpressure with detonation.
associated

The large volume dump-vesselwas hydraulically testedat 11.25baro. In accordance

with the relevantregulationfor the design of large pressurevessels,a pressurerelief
valve wasfitted. The openingpressureof this valve was setat the designpressureof the
vessel.

2.6.2 Transmissionof the explosionto auxiliary equipment

The preparationof fuel-air mixture involved auxiliary equipment such as external

pumpsandmonitoringsystemswhich all specificationwill be describedin later section.

A safety operation procedurewas designed to ensure that all procedures will be
followed and applied for each test. In order to isolate the test vesseljust prior to
ignition, variousball-valveswereclosedand fuel-lines disconnected.The sparkignition
circuit incorporatedsafety features that involved the closure of several electrical
contacts.This is describedin detail in later section.An importantpart of the circuit was
a mixture inlet line safetyinterlock. The filling line had to be disconnectedfrom the test
vesseland thenconnectedto the electricalinterlock beforeignition took place.The gas
cylinder bottle wasthereforeisolatedfrom the vesselbeforeignition.
 41

2.6.3 Releaseof combustiblegasinto the Test room

This hazardwasnot specifiedto this facility but was also applicableto any University
areawhere combustiblegaseswere presentand handled.In the event of an accidental
explosionwithin the Test room, it was expectedthat the blast would be relievedto the
outside through the large window along the external wall. This large vent area was
expectedto reducethe pressurebelow level that might damagethe partitioning walls
that isolatedthe Control room.The glasswindows werecoveredwith a clear membrane
that wasspeciallydesignedto limit the glasspigmentation.

2.7 Operating procedures

Due to largenumberof operationsandsafetychecksinvolved in a single explosiontest,

an OperatingProcedureform was devisedthat had to be completedduring eachtest.
This form containsall experimentalstepsthat neededto be followed as well as the
safety procedures.A flow diagram is shown in Fig. 2.8 that illustrates the safety
proceduresand guidelines that account for any deviations from the operational
procedure.As part of the experimentalprogramme,threerepeattestswere performedat
each condition and these demonstratedgood reproducibility, with peak pressures
varyingby lessthan ±5% in magnitude.

2.7.1 Fuel entrainment

In order to create a combustible mixture of known composition, the volume of both fuel

and air introduced into the test vessel had to be known accurately. The theory of partial
pressure was applied to an accuracy of 0.1 mbar (0.01% of composition). The final pre-
ignition mixture pressure was set at ambient pressure (- I atm) for
all tests. For
example, to make up a mixture with 10% fuel/air ratio, the test vessel was evacuated to

a vacuumpressurethat allowed fuel to be introducedbefore letting the ambient air to

fill in the test vesselto reachambientpressure.
 42

2.7.2 Ignition procedure

All isolating valveswere closedand the mixture inlet line was disconnectedfrom the
test vesselandconnectedto a safetyinterlock.The dataacquisitionsoftwarewas armed
and ignition circuit integrity was indicatedby light on the ignition control panel after
the openingof the gatevalve.The 'fire' button waspressedand this activatedthe spark
plug and by this, it triggeredthe data collection system.Once the samplingtime was
over, the raw transducersdata weredisplayedusing FAMOS software.Output signals
were recordedto check whether a pressurerise had occurredin the system and the
maximumoverpressure is not exceededthe critical adiabaticpressureof the fuel/air.

2.7.3 Vesselpurging

Purging involved evacuationof both test vesseland the dump vessel using vacuum
pump B (refer to later section).Once a sufficiently low vacuumpressurewas attained,
air was allowed to enter the systemfor more than ten minutes.This procedurewas to
ensurethat the systematmospherematchedthat of the outside ambient air. The test
facility wasthenreadyfor the next testpreparation.

2.8 Ancillary equipment

The preparationand ignition of fuel/air mixture inside the test vesseland its subsequent
purging involved various equipmentswith associatedpipe works and valves. As
described in the previous section (refer to section 2.7), the procedures involved
including the evacuationof the test vessel,mixture preparationand recirculation and
control of the dump vesselpressure.Isolation of all ancillary equipmentwas required
beforeignition andthe test vesselhadto be isolatedfrom the connectedpipe (if venting
with relief pipe geometrywas carriedout) anddump vesselduring mixture preparation.
 43

2.8.1 Isolation valvesand pipe works

All valvesthat attachedto the test vesselandthe dump-vesselare neededto be closed

prior to the ignition in order to stimulate the confined vessel condition. The fuel

connectingpipe, the pipe attachedto the vacuumpump and the pipe attachedto the
barometershould be disconnectedprior to ignition. Step-by-steptest procedure is
shownin Fig 2.9.

2.8.2 Mixture pressure monitoring system

The theory of partial pressurewas usedto preparea mixture of an accuratelyknown

composition. This required accurate knowledge of the test vessel pressure under
vacuum.A Diametric type 600 Barocel pressuresensorwas used to monitor the test
during
vesselpressure The
mixture preparation. unit was fitted into the test vesselfilling
line circuit so that the mixture pressurecould be monitoredat all times. Its principle of
operationwasto transducerabsolutevacuumpressureinto a dc output voltageprecisely
proportionalto input pressure.Advanceddesignallowed pressuremeasurementover a
wide rangeof input pressurewith high accuracyandstability. The unit was connectedto
a Diametrics type 1500 digital pressuredisplay. The combined system allowed the
mixture pressureto be monitoredto a resolutionof ±0.05 mbar.For a final pre-ignition
pressureof -I atm,a 10%CH4-airmixture could be preparedto an accuracyof 0.05%.
 44

Evacuatetest and dumpvessel

Is vacuumpressurestable?
(Lessthan0.2 mbar/minleakrate)

Yes
No

Allow ambient air Calculatethe fuel required to be filled

into the systemand into the test vessel using partial

abandontest pressuremethod

Does pressure exceed the target fuel

pressure? I
Yes

Fill up the ambientair into the test vessel

until it'reaches ambient pressure. Allow
No
time for a stablepressurereading

Close fuel valves and Yes

isolatethe fuel cylinder. Disconnectfilling line andfit mixture inlet line

Follow nitrogenpurging to safety interlock. Close all isolation valves.
Checkpowerto ignition system
Checkmain fuel cylinder valve is closed.

Yes

All personnelgo into the Control Room

Checkall lights on control panelare ON
i. e. Door locked
Door latched

Figure 2.8 Flow diagram illustrating safety and operational procedure

 45

Arm datalogging system

PressMulti-Spark fire button

x
Yes

Did data logging system

No
trigger?
I
Is this thefirst ignition loop?
Yes
No
Yes Savedata

Follow nitrogen Yes

purging Repeat igniti( as a pressure rise
III--
.orded?

I
Yes

Go into the test room and

follow the purgingprocedure

Figure 2.8 contd.

 40

6-1

,. I --d by
ý;

N,, . ......... .
..... ........... .... ... .. Ppll

1--l. --. 1 -,
01, - VW, V: i K V4 50",

Figure 2.9 Example ofoperating procedure for carrying explosion tests.

 47

2.8.3 Vacuum gate valve

For all testvessels,a series12VAT vacuumgatevalvewith a nominalopendiameterof

0.162 m was used to isolate the test vesselfrom the connectedpipe and the dump

vessel.It consistedof a light-weight aluminium body and gate with Viton seals.Shaft-
feed-throughwas employedwith pneumaticactuation.Valve operationwas controlled
by a solenoidvalvein the compressed
air supplyline pipe.

2.8.4 Evacuation system

As mentionedearlier,mixture preparationin test vesselI and 3 involved evacuationof

test vesselto certain vacuumpressure.The large vacuum pump B was used for this
procedureas it involved bigger volume.For a smallerTest vessel2, a smallerpump A
wasusedfor thesamepurpose.

2.8.4.1 Vacuum pump A

This was an EdwardsEIM18 singledirect drive, rotary vacuumpump. It hasa nominal

displacementrating of 340 I/min. the pumping mechanism was of the slotted

rotor/slidingvanetype. Direct drive was providedvia a flexible coupling from a totally

closedfan-cooledmotor. It was controlled by an on/off switch locatedon the mixture
preparationpanelin Testroom.

2.8.4.2 Vacuum pump B

This wasan EdwardsE2M175to-stagerotary vacuumpump. It wasan oil-seatedpump

designedfor reliable,long-termoperationin both laboratoryand industrial environment.
It hasa nominal displacementrating of 2967 I/min. the pumping mechanismwas of the
sliding vanetype with high and low vacuumrotor and statorassembly.Direct drive was
provided through a flexible coupling by a four-pole, three-phasemotor to EP54
enclosurerating. The pump was water-cooled.Water flow was controlled by an
electrical isolation valve. Power to the pump was via a main isolation valve and soft-
starter.
 48

2.8.5 Ignition system

A single conventionalcar sparkplug was usedto ignite the fuel-air mixture in the test
vessels.Ignition was actuatedfrom a standardcombustionengine spark (16 J) via a
high-capacitancedischargecircuit. The circuit diagramis included the various safety
interlocks. The system ensuredthat the following operationshad to be carried out
before spark ignition could be effected.

The gate valve used to isolate the test vesselsto pipe and dump vessel during the

mixture prr.paration had to be opened prior to ignition. This action to ensure a

vented explosion and avoid the high pressure attained inside the test vessel
resultedfrom a closed-vessel
explosion.
The connectingdoor betweenthe Control room and the Test room had to be

closedandlocked.
0 The mixture inlet line had to be disconnectedfrom the test vessel. This
effectively meantisolationof the fuel supplyprior to ignition.

Indication lights, fitted to the ignition panel at the Control room were monitored to
check whether this action had been carried out. The data acquisition system was
connectedto the ignition circuit for triggeringthe fire button for datacapture.

2.8.6 Nitrogen purging system

If the fuel-air mixture was not ignited might be due to the failure of the ignition spark
plug or electrical faulty in ignition circuit, a relatively large volume of explosive
mixture would remain inside the test vessel.Before any repairscould be made,the test
vesselneededto be purgedof the gasmixture. A nitrogenpurging systemis one of the
special case circumstancesthat neededto be followed if this case happened.This
procedurecarriedout by injecting high pressurenitrogenthat would mix with the fuel.
A standardpressuregaugeon the Barocel could be monitored so that the volume of
 49

nitrogen introduced was sufficient to inert the fuel-air mixtures. At the same time,
simultaneousoperation of vacuum pump B allowed the mixture to be-safety purged and

releasedto the atmosphereoutsidethe Testroom.

 CHAPTER3

VENTED GAS EXPLOSIONS: THEORY AND

CRITICAL REVIEW
 50

3.0 Introduction

The ventingtechniqueis a popular andeffectivepreventionmethodto reduceexplosion

hazardin industrial containersof flammable gases,liquids and powders (Eckhoff,
1991).Venting is possiblebecausemost hydrocarbon-airmixtures have fundamental
burning velocity, S. normally less than I m/s although flame speeds may be
considerablyhigher while the velocity of soundin air is around 340 M/s at normal
temperatureand pressure.This meansthat, pressuretransmissionmay be regardedas
but the rate of pressurerise will be relatively slow, subjecting
effectivelyinstantaneous
the structureto be protecteduniformly to a stressfixed by the amount of combustible
material. Vent therefore may be located with equal value wherever feasible in a
compartmentor duct, providing the run-up distancesfrom the point of ignition to the
ventis similar (Anthony, 1978).

The implementationof venting has motivated several theoretical and experimental

studies,the resultsof which haveinspiredthe developmentof engineeringstandardsi. e.

NFPA 68 (NFPA68,2002) and European Guide on venting (2007). The
and guidelinesprovidedby thesepublicationsare basicallyregardson
recommendation
the question of appropriatescaling of the correlation parametersas the analytical
are invariably usedto predict the vent arearequirementsfor conditions
representations
that arebeyondthosecoveredby the experimentswhich supportthe designmethod.

Essentially,four quantitiesneedto be known beforea properestimationof the vent area

canbe made(Lunn, 1992).Theseare:
The reducedexplosionpressure,Pred.This is the explosionpressurethat should
not be exceededand which dependson the strength of the vessel. This is
generallydesignatedas Predand has a unit of bar. Predshould not exceedtwo
thirds of the burstingpressure,P,. It may be that the strengthof the vesselis not
accurately known and such a lack of information must be an important
considerationwhendecidingwhich methodof vent areacalculationto use.
The vesselvolume, V. Most of the basic methodsfor estimatingvent areasare
limited to compactenclosureswith lengthto diameter(IJD) ratio lessthan 5.
 51

The mixture reactivity in terms of gas deflagration index, KG or dust

deflagrationindex, KST or burning velocity, &- It can be shown later the
between
relationship KG S,,
and and how KG is dependent on vesselvolume.
Locationof the ignition sourceandthe turbulentfactor. It had beendepictedthat
endignition gave higher Pred
as the flame haslongertravel distancecomparedto
central ignition. Turbulencegenerated through the interaction of the gas flow

aheadof the flame with obstacleinside the This

vessel. is normally been taken
into accountby a turbulentfactor, 0 wherethe turbulentburning velocity, STiS
in termof

ST=I +,8S.

3.1 Parameterinvolved in empirical equations.

In venting,the pressuredevelopedis dependentupon the vent coefficient, K=VW /Av

which expresses the effect of the vent openingareaand the burning velocity, S,,of the
gas-airmixture. However this parametersare not be used as straight forward. This

concepthas been applied by Cubbage and Simmonds (Cubbage and Simmonds, 1955,
Cubbageand Simmonds,1957),Cubbageand Marshal(Cubbageand Marshall, 1974)
and Rasbash
(Rasbash,
Drysdaleand Kemp, 1976).Bartknecht(Bartknecht,1993) used
K terms to expressthe vent flow pressureloss and vent bursting pressurein his
KO to the
correlationand value express gas/air mixturereactivity. KG is defined as

Kc,=(dp) V 1/3 bar-m/s (3.2)

dt
max

where dP/dt is the rate of pressurerise (bar/s) and V is the vesselvolume (M3). Kr, is
given as 50,100,140 and 550 bar-m/s for methane, propane, town's gas and hydrogen
in 5 litre sphericalvesseldoneby Bartknecht(Bartknecht,1993).Thesevaluescan only
be used with the Bartknecht's correlation as KG value is not to be constant but to
increasewith vesselvolumeasbeeninvestigatedby Chippet(Chippett, 1984).Thus, KG

value should not be assumedconstant and yet, be regardedto be representingthe

mixture reactivity.However,the useof burning velocity, Sugivesthe sameproblem. In
venting,it is assumedthat Suis constantthroughoutthe explosiondevelopmentso does
 52

the flame speeds.This assumption is not valid as the higher compression pressure and
temperature of the unburnt gases changes the burning velocity from that of the initial

gaspressureand temperature. However, the change of S,,on theseparameters is a small

effect (-20% increasein S. at the end of the explosion),Further,S,,is alsodependentto
the mixture reactivity,addingto the difficulties in quantifyingthe parameter.However,
the assumptionof constantS,,is not a bad approximationespeciallywhen Suis taken as
a little higherthanthe valueat normalpressureandtemperature.

A more completedtheory of closedvesselflame propagationand the derivation of the

relationshipbetweenrate of pressurerise and burning velocity, Su was presentedby
Nagy et al (NagyandVerakis, 1983).

2/3y I/Y]2/3
dP 3Su ýpm 1/y_pl/r)1/3 Po 8)
Tt (P
-R *p (2-11Y - 6) '00p 1_ý p(3-2ly-, (3.3)
01

From Eq. 3.3, onecaneasily derive the dependentof Ko (from Eq.3.2) to S,,.Kumar at
al (Kumar, Bowlesand Mintz, 1992) also derived an equationfor KG as a function of
S.. This equationalso assumesthat S,,is the constantand doesnot vary with radius as
well aspressureandtemperature.The equationstatesas,

KG = 4.84S. (P,. / PI)"r(P», - Pl) (3.4)

1
Further,using an assumptionof only little pressurerise beenexperiencedin the closed
sphericalvesselexplosionin the first half of the flame travel andachievedits maximum
about98 % of adiabaticpressurerise in the last half of the travel given KG valuein term
of,

KG = 3. I6(P. - Pi)Sý,E (3.5)

This formulation derived by Andrew (Andrews, 2004) with regards on the constant
burning velocity through the flame travel. From Eq. 3.3 3.5, it is clear that the
-
dependantof dP/dton S,,.From this relationship,it can be seenthat KC1= constantS,,.In
dust explosion,the relationshipof K., andturbulent burning velocity, STexistedand this
can be relatedwith laminar burning velocity for dust i.e. ST= PSuas shown in Eq. 3.1.
 53

Thesetwo parametersare mainly usedto correlatethe vent design by the researchers

evennot in directorderin orderto meetminimum deficiencyfrom experimentalresults.

3.2 Venting mechanism.

Experimentalresults or calculationsfor vented explosionsare usually presentedby

expressinga term containing the peak (reduced) pressure,Predas a function of a
parameterdescribesthe amountof venting. In contrastto the closedvessel,explosions
in a ventedvesselare characterisedby the maximum reducedexplosionoverpressure,
P. d insteadof the maximumexplosionoverpressure,P.. (shownin Figure 3.1) and by
the maximumreducedrate of pressurerise, (dP/dt)redinsteadof the maximum rate of
pressurerise,(dP/dt)max
(Siwek, 1996).The reducedexplosionpressureor Pedis usually
designedto be approximatelytwo-thirds of the pressurerequiredto rupture the vessel.
In a given vessel,the 'reducedexplosion pressure'will dependupon various factors

of the vessel,numberand location of the vent, the openingvent

suchas the size/shape
pressureand inertiaof the vent cover,the presenceof obstructionsinside the vessel,the
of the gasor dustetc.
explosivecharacteristics

P,
LD

a
Time (s)
Figure 3.1 Pressurebehaviorversustime for a closedandventedexplosions.
 54

The basic principal of explosion venting provides for the rapid opening of a vent of

sufficient area to allow unburned gas and explosion products to escape,thus limiting the
pressurerise to an acceptablelevel. On the general basis, the reduced pressure attained
in the vessel before venting must be less than the mechanical strength pressure of the

containers. The vent opening must be correctly sized to allow the expanding gasesto be
vented at a rapid rate so as to limit the internal development of pressure.The acceptable
pressurerise is determined by the requirement that the vessel should not rupture and in
some cases,it should not deform. Essentially, a vented explosion can be considered to
have three phasesas shown in Figure 3.2:

41 An initial confined explosionphasewhich occursprior to removal of the vent

cover
At this stage, the rate of pressurerise will be determinedby the rate of
productionof burnt gaseswhich in turn dependsupon the flame speed(Marked
by a). Theoretically,flame will expandsphericallybefore it distorted to semi-
hemisphericalshapeand stretchtowardsthe vent. The flame areais increasing
and hence,the rate of massburnt increases.The increasingof massburnt rate
will increasethe flame speedand thus, higher dP/dt as well as the shortertime
taken to reach a given pressure.It means that the first requirement of an
effectiveexplosionrelief is that an openvent shouldbe createdat an early stage
in the explosionprocess.By this means,the vent cover should have as low
breakingpressureaspossible.

A vent removalphase.A situation when an open areais createdthrough which

gasescanescape.In industrialpractical,a vent cover is neededto sealthe vessel
in order to preventthe escapeof fumes or combustionproductsbut is designed
to fail at a low pressureand allow an outflow of gasesto be establishedat an
early stage.The transition betweenthe confined explosionphaseand the vent
removal characterisedby P, in Figure 3.2. After this phase, a significant
decreasein explosionoverpressureshouldresult but sincethe flame front is still
expanding,the rate of pressurerise, dP/dt is increasingandthe pressurewill rise.
In order to minimise the pressure developed during venting, the breaking

pressure,P, should be low as possible.Lower breakingpressureleadsto small

flame elongation and hence, smaller flame areas and as a result lower Pred
 55

whereasat higher breakingpressure,the effect of flame distortion is reducedas

it takeslonger time to break and hencewould result on higher Pred (McCann,
Thomasand Edwards, 1985).It should be noted that in higher P,, the flame
becomecellular prior to removalof the relief panel.

A venting phase. A phase during which gas flow out of the open vent is
established.In this case,the pressurewill riseagainafter the vent cover hasbeen
removeduntil the flame front reachesit maximumareawhich denotesas P2 in
Figure 3.2. Cubbageand Simmonds(Cubbageand Simmonds,1955, Cubbage
and Simmonds, 1957) thoroughly investigated the development Of Predin
industrial ovens using town gas/air and methane/airmixtures by varying the
location of the vents and different breaking pressure of vent covers. The
significantresult from their work was the doublepeakpressureprofile attained.
Doublepeakpressure-timeprofile occurswith low breakingvent pressurewhilst
single pressurepeak profile occurs with vent cover having a high breaking
pressure(Harris, 1983).A higher rise in vesselpressureis observedfor lower
venting pressuresand this is explained by increasedflame area and a longer
period of unburned gas venting, which is less efficacious than burned gas
venting (Bradley andMitcheson,1978b,Harris and Briscoe, 1967).Effectively,
the total surface area of the flame front is greatly enhancedby the sudden
openingof a vent, and this in turn increasesthe overall combustionrate of gases
within the vessel.Thus, very strong secondpressurepeaks are often detected
following the initial pressuredrop whenthe vent opens.

Turbulence and stretching of the flame surface towards the vent opening can also

result in an increase in the combustion rate (Buckland, 1980). The turbulence may
be caused by increased gas flow velocities and by the opening of vent covers.
Burgoyne and Wilson (Burgoyne and Wilson, 1960) have demonstrated that, all

other things being equal, a vent which opens smoothly can give a lower Predthan
does a diaphragm which burst abruptly at the same opening pressure. The
very
presenceof a vent cover can increaseturbulence in an explosion,

A further peak,P3might occur if the gascools down sufficiently

rapid to reducethe
pressurein the vesselbelow ambient,causingre-entryof ejectedunburntgases.The
 56

external explosion that gives rise to the P3 is due to the fast flame propagation in the
turbulent external vented premixed gases. This pressure increases as the vent area
decreases or high K value as this will produce higher and turbulent jet velocity out

from the vessel. Usually P3 develops because the size and shape of the enclosures,

the position of ignition source as well as the stretching and movement of the flame

towards the vent as it opens, allowing burnt gas to be vented while combustion is

still taking place inside. Transient pressure may also be affected by oscillation being

set up in the system (Anthony, 1978) as represented by P4. This oscillatory type

combustion or Helmholtz oscillation will further increases the rate of combustion

which that increase the internal pressure inside the vessel and this oscillation

pressure gradually damped out as the flame expands. However, the burning rate
during this phase is enhanced by the turbulence generated in the shear layer between

the outflow burnt gas and unburnt gases within the vessel (Cooper, Fairweather and
Tite, 1986). The strong interaction of the flame front, the shock/pressure wave and

physical back flow into the vessel generally enhance the rate of combustion inside
the vessel and further induced Taylor instability on the section of flame front surface
farthest away from the vent. As the flame is continued to expand, it eventually

encounters the walls of the vessel and this will suddenly decrease the flame area and
hence the rate of pressure rise. Work by Solberg et al (Solberg, Pappas and
Skramstad, 1980) confirmed the existence of third pressure peaks assembling the P4

trace in 35m 3 vessel in which the vent were initially covered. They found that a

strong flow field will be generated at the opening of the vent which the flame will
be strongly accelerated towards the vent and Taylor instability may develop at the
worst case of central ignition.
 57

1ýr.

10

ci
m
E to

tA
vi
i2 4G

CD 21

-S
TIME/ s

Figure 3.2 The pressure-timevariation in a ventedexplosion.Graph reproducedfrom

Cooperet al (Cooper,FairweatherandTite, 1986)

The relative size of peaks P, and P2 are determined by the size of the vent relative to the

vessel, the magnitude of P,, the flame speed and the scale and intensity of turbulence

when the vent breaks. As the vent area of the vessel or enclosure is smaller, the pressure

required for vent cover removal is increased. It is found that the bursting vent pressure,
P, is inversely proportional to the area of the vent cover (Cubbage and Marshall, 1974).
Further, the smaller the A, i. e. high K, the smaller is the vent area available upon its

removal and the less rapid, therefore, the egress of combustion products were impeded.
Due to inertia of the vent cover, a finite time is required immediately after P, is reached

to totally remove the vent cover sufficiently far from the enclosure for the flow of gases

out of the enclosure and thus, the weight per unit area of the vent cover is another
important factor influencing Pred.The heavier vent cover is not as effective as the lighter

material in minimising the pressure developed (Cubbage and Marshall, 1974, Molkov,
Grigorash, Eber, Tamanini and Dobashi, 2004).

An increase in P, will cause P, to increase with respect to P2; while high flame speed

and high turbulence will cause P2 to increase with respect to Pl- P2 also increases with

respect of P, as the vent size becomes smaller for a given vessel (Anthony, 1978).
Further, at certain condition, there is only one peak is
observed. The merging Of P2 with
P, corresponds to a situation where dP/dt becomes
negative after the operation of the
vent which occurs when the vent is relatively large or flame speed is low. However, the
 58

merging of P, with P2 correspondsto the situation wheredP/dt remainspositive after

the openingof the vent and this situation generallyoccurswhen the vent is relatively
small andflamespeedsare higher.

Thus the explosionpressuredevelopedin ventedexplosiondependson threequantities,

the pressurerequiredfor vent to release,P, the back pressuredue to inertia of the vent
cover andthe backpressurecausedby restrictionto the gasflow throughthe vent and to
these, the turbulent parametermust be added. These effects on Predwill become
complicatedfunction as other variablesfrom thoseeffects mentionedrelated to other
dependentvariablesand all parametersneed to be accountedin order to perform the
appropriatedesignfor venting.

3.3 Venting theory

For the purposeof correlating the theory and experimental results, P2 will be assumed as

the dominant overpressure as P, is usually regarded as the bursting burst pressure. The
problem arises as most investigators of vented explosions do not report the pressure-
time profile but simply report the maximum overpressurethey attained in that particular

experiments. Hence, it is normally not possible to know whether this is caused by PI, P2
or P3.

However, understandthe physics of vented explosion which regardsthe flow of the

unburntgasesaheadof the flame ventedout throughthe vent allows overpressureto be
calculated.The procedurewould allow the form of correlationwhich includes P2to be
deducedand further, will then be comparedto publishedexperimentaldata in vented
explosions. Thus it will be assumedthat in most cases of vented explosion the
maximum P,,d is causedby the flow of unburnt gasesthroughthe vent and hence,P2
shouldbe takenas maximumoverpressurebasedon the assumptionmade.

The mostusedandrecommendedcorrelationfor venting gasdeflagrationis

adoptedby
NFPA 68 (NFPA68,2002) andEuropeanGas Venting Guidance(2007)rely the
on vent
correlationdevelopedby Bartknecht(Bartknecht,1993)for high-strengthenclosureand
Swift's correlation (Swift, 1983) for low-strength enclosures,
respectively. The
 59

correlations apply for a compact vessel which Bartknecht defined as those with LJD <2
although the draft European Standard (2007) has applied the correlations for UD < 3.
Bartknecht'scorrelation(Bartknecht,1993)is given asthe vent area,A, as a function
the vessel volume, V2/3 multiplied by a complex term that includes the mixture
reactivity KG,reducedoverpressurein the ventedexplosion,Predand static vent burst
pressure,P, The V213dependenceof overpressureon the test vessel volume is a
characteristicof spherical or compact vessel explosions,where the flame remains
mostly in sphericalshapeduring ventingprocess.If the sphericalflame propagatdsat a
constantrate irrespectiveof the vesselvolume, Predshouldbe only dependenceon Av
andV213or K. Bartknechtexpressedhis correlationas:

I
0.175(Pstat-0-1) VY3
-0.0567 (3.6)
=[0.1265logloKc,
pl 182 0172
p j
,.,

where Kr, is deflagrationindex (bar-m/s),P, is the static burst pressure(barg), V is

volume (M3). Equation3.6 is a dimensionalrelationshipwhich yields A, as a function
of two termsi.e. the vent flow pressureloss for the first andthe vent bursting pressure
term for latter which gavea linear influenceof P, on A, Thereis no fundamentalbasic
for Eq. 3.6 as it is a pure empirical correlationand relies on appropriateexperimental
data to determinethe four adjustableparameterswhere the detailsof the data analysis
have not beenmade public (Bartknecht, 1993). Again, the correlation developedby
Bartknecht(Bartknecht,1993)can be saidempirical asthe dependencyof KG with S. as

mentionedearlier.In his work, Bartknechtcarried out ventingexplosionmainly in four

different fuels which are methane,propane,town's gas and hydrogenat volume range
betweenI to 60 m3.If the equationis appliedto a standardvent burst pressureof 100
mbar where the influence of P, and weight of inertia of the vent, w can be assumed
negligible,Eq. 3.6 cdnbe transfomed into:

1 0.1265log K. - 0.0567 constant

K P,, 0,582 pl-OIT2
(3.7)
d

Given a valueof Ko for eachgas'as55 for methane,100for propane,140for town's gas

and 550 for hydrogen will give the constant as 0,164,0.20,0.212 and 0.290
respectively. As been discussedearlier, Kr, increaseswith vessel volume but yet,
Bartknechthasnot beendeterminedthe Ko value at different
vesselvolume.
 60

(bar] -

0
2,0 - ------------------ --- - --------------------------

\OON1, o\
1,0 - ---- ---------------
0.8 -
0,6 -
0 v=i rn,
3ý-
0-
--------------------------
ErgebnIsse:

CL Vý 2 M"

0,4
0
A V-10
0 V=30
m
m3
U, V-60 m3
%
noch kub'ischem Gesetz cl
0,2 euf 1 rri3umgerechnet 0\

['m 21
0,02 0,040,06 0.1 o,2 0,4 0,6 0,8
O,OB F

(bar)

Bartknecht

2,0 -
correlation
------------------- -----------------
line

1,0 -------------------------- --------------------

0,8 Ergebnisse:
0 *'\, +
C V- 1 M3
V= 2 m3 =30AI
0.0,6 0
A V=10 m3+
0,4 0 V=30 m,
C3 V-60 m
nach kubischem Gesetz 0a \z,
0.2 cuf 1 rrýumgerechnet *0\+

0.11 111! II
ýII-ILI. 0,2 0.4 0.6 0,8(M"]
0,01 0.02 0,04 0.06 0.1
0,0a 1/K
Figure 3.3 Bartknechtdata for propane-airat different volumesat E= 10J,P, = 100

mbars (Bartknecht,1993).* is for at 30m 3 on I/K. + is for at

propane/air V
methane-air
3.Reproducedfrom Andrews(Andrews,2004).
=2m

The correlationdevelopedby Bartknecht(Bartknecht,1993)can only be appliedon its

own limitation asthe equationis the empiricalderived from his experimentalworks and
 61

no other publishedventing data includedas notedby Siwek (Siwek, 1996)in his study
on publishedcorrelationsin ventedgasanddust/air-mixtures.

Figure 3.3 showsthe original data from Bartknechtfor propane/airat different volume.
From the figure, Bartknechtportrayedhis correlation basedon the experimentaldata
obtained for V= 10 m3- It suggestedthat the correlation developedby Bartknecht
(Bartknecht,1993)was purely empirical with no safetyfactor offeredfor ventingdesign
At it
purposes. the x-axis, representedthe vent area, denoted as F (m2) and Pm for y-
axis. However,when the result data is re plotted for V=2m3, representedby the (+)
sign, it is found that only one data agreeswith the line that drawn from the correlation
line. After re-analysed,it is found that F parameterat x-axis is a mistakeand it should
be correctedto I/K. To justify the alterations,the methaneexperimentaldata,denoted
as * sign from Bartknecht'swork for methane/airat V= 30 m3 were re plotted and it
showeda poor agreementwith the correlationline drawn. From the figure, it is implied
that the same vent area is required irrespective of the vessel's volume which is
obviously incorrect.Further,the resultsimplied that lower overpressureattainedfor the
sameK in the largestvolumei.e. 60 m3

From Fig. 3.3, the overpressuresare increasingwith volume for vesselvolume of I to

10 M3 but decreasingfor higher volumes.The self-accelerationmechanismcan be the

reasonableexplanationfor the increaseoverpressurefor the first condition but there is

no physical explanationfor the latter case.As will be discussedlater, the volume effect
is not correctlycorrelatedby the K parameterin Bartknecht'scorrelation.

NFPA 68 adoptedSwifts equationto apply for design and scale-upof venting system
purposes and it seem that the results obtained were in good agreementswith
experimentaldataobtainedby Donat (Donat, 1977)and Harris and Briscoe (Harris and
Briscoe, 1967). Again, there is a limitation in using this approach as it can be applied on

certain conditions i. e. Predis not more than 200 mbar and only suits to low-strength
enclosures.

The Swift correlationhasa form of.

 62

CA,
A, = (3.8)
Pled 0.5

where C is the deflagration constant (kPa)"2 that determined by experiment or

calculationenvelopedthe gasreactivity terms,dischargecoefficient (typically 0.7) and
turbulencefactor (typically 5) and A, is the vessel surface area (M) . Further, the
relationshipbetweenA., and V2/3can be definedas As = constantV2/3whereconstantis
equalto 4.84 for sphere,6 for cubic, 6.3 for rectangularand 5.81 for cylinder with L/D
= 2. In K term, Eq. 3.8 can be transformedinto:

I= constant
(3.9)
K

The use of A, insteadof V2/3by Swift is followed by Bradley and Mitheson (Bradley
and Mitcheson, 1978a,Bradley and Mitcheson, 1978b) as Swift predicts that a larger
vent areais requiredfor cubic vesselrather than sphericalvesselat the samePredand
volumewhich is not in the caseof Bartknecht'scorrelation.This meansthat A, is the
key parameter to include the influence of the vessel geometry/shapeson the
The simplicity and easeof useof Eq. 3.9 is questioned,underestimatethe
overpressure.
complexity of large-scalecombustionphenomenaand did not take into consideration
the effect of vent openingpressure.Swift (Swift, 1983)basedhis design equationon
theoretical development that indicates the independence Of PredOf the vent opening

pressure,if the turbulenceenhancementitself is also independentof the vent opening

pressure.The turbulenceinducedby the vent openingor the flow of unburntgasesover
and around the obstaclesare not yet realistically be demonstratedespecially in large
scaleconditionsasthe previousinvestigatorsput constantto give effect on turbulenceto
fit with their experimentaldata andthesevalueswere differed from othersand still, this

root problemis not yet resolved.

The mostclassicand often referencedstudieson ventedexplosionvent designhas been

published by Bradley and Mitheson (Bradley and Mitcheson, 1978a, Bradley and
Mitcheson,1978b).They madea detailedcomparisonof their theoreticalresultswith an
extensivecompilationof the experimentaldata.Their presentationof the model results
was in a form of plots for dimensional (in atmospheres)reducedexplosion pressure,
 63

APred : ý/-S
versusa dimensionlessvent parameter, asfollows for initially closedventing
Le, with the ventcoveron place;

Pl, /S)-0,6993
d ý 2.43(A for P 2ý1barg (3.10)

P = 12.46(A/ S)-2for P 51 barg (3.11)

Cd AC,, (p,,
with j (3.12)
s A$Su A

C,, is
where i the speedof sound(m/s), p, is the unburntgas density 3)
(kg/m and Pbis the
burnt gas density (kg/m3) The dimensionlessvent parameterhas been termed by
.
Molkov (Molkov, 1995, Molkov, 2001) as turbulent Bradley number, Brt where the
only different from the original equationis the addeddeflagration-outflow-interaction
number,X/g.The model suits ideally for sphericalvesselflame propagationas it is the
basisfor the equationsdeveloped.

If the Eq. 3.10 and 3.11 changedto K term as followed in Bartknecht and Swift's
equationabove,it gives,

1 0.28 (E - 1) contant
CdCuipred
, 1.43
= -ýred
1 43 (3.13)
K .

1=3.53S (E - 1) contant
,= Pud 0.5 (3.14)
Kc dC. ip, ýd0-5

The equationdevelopedby Bradley and Mitheson (Bradley and Mitcheson, 1978a,

Bradley andMitcheson, 1978b)assumedthat the pressurerise will be enhancedby the

venting of the unburnt gasesrather than the burnt gasesand this is taken as a 'worst
case' scenariofor a safe vent design. The pressurereduction enhancedby burnt gas
venting is supportedby Maisey (Maisey, 1965), Rasbashand Rogowski (Rasbash,
1986)andYao (Yao, 1974).

Nevertheless,this approachis been questionedif there was a significant P, (Lunn,

ICheme,1984and 1990).The expandingflame front which associatedwith substantial
 64

gas flow aheadof it and the presenceof openingvent alters the flow patterns.These
conditionsgenerateintenseturbulencetowardsthe flame propagationin which increase
the burning velocity and hencethe maximum overpressure.
At this particularcondition
where the postulateddiscrepancybetween the theory and experimentalresults in a
presenceof turbulenceassociatedwith the increasein burning velocity, Bradley and
Mitheson (Bradley and Mitcheson, 1978a, Bradley and Mitcheson, 1978b)
recommendedthat a five fold increasein burning velocity,abovethat of the laminar
value would benecessaryto fit the experimentaldata.

The recentalternativevent designcorrelationat increasedpressureandtemperaturethat

is adoptedby NFPA 68 and draft of the EuropeanVenting has been introducedby
Molkov et al (Molkov, 1995, Molkov, 1999, Molkov, 2001, Molkov, Baratov and
Korolchenko, 1993,Molkov, Dobashi,Suzuki and Hirano, 1999,Molkov, Grigorash,
Eber and Makarov, 2004, Molkov, Grigorash, Eber, Tamanini and Dobashi, 2004,
Molkov, Korolchenkoand Alexandrov, 1997). They had called the correlationsas a
&universalcorrelation'for prediction of reducedpressureand vent area.It is basedon
the modified correlation from the original correlation developed in 1993 and
1995.(Molkov, 1995,Molkov, BaratovandKorolchenko,1993)

In the correlation,deflagration-outflow-interaction
numberor can be called turbulence-
factor,Z1,U and Bradley-Mitchesonnumber, Br are introduced where the so-called

deflagration-outflow-interactionnumber Zlu was derived by fitting the calculated

pressure-timecurvesto the experimentaldata from variousauthor (Razusand Krause,
2001). The parameter represents the turbulent generated during the explosion
development.TheBradley numberandthe deflagration-outflowinteractionarefollows:

; r,, d Br, -"4 for ; r,,

= d <

;r,,d=7-6Br, "' for; r,,, 1 (3.16)

d>

Br, = Br ýu (3.17)
V3
-6; r z
 65

A, c",
Br = V2/3 -
S,,(E - 1)

0.37
x (1+10V"')(l+O. 5Br) 1
0.9 (3.19)
ß11+x,

P,(bar absolute)
Where TrV=- Y is the specific heat
is the expansionratio Of Pu/Pb,
Pi ,E

ratio for unburnedmixture,c,ýjis the speedof sound (m/s) and R. is 'pi' number. As the
the
samewith othercorrelations, original equations (Eq. 3.15 and 3.16) will be changed
to K term. This givesas,

V3-
-6z S,,(E - 1) 1 constant (3.20)
TT
K Cid Pred 0.4 Pred0.417

2= 2
I= V3--6z 7- P,,
X S,,(E - 1) 7-P,, d d
(3.21)
JE-1y', u' constant
K C,,, 66

The equationis valid for unobstructedenclosuresbut the authoracknowledgesits use

for congestedenclosuresas well. The first attempt to quantify the level of turbulence
in
generated ventingexplosions has beenmade by Munday (Munday, 1963)with useof
detailedmodels,experimentsandbestfit approachand from then, the works to include
turbulencefactor Nevertheless,
are progressing. the value of turbulencefactor is based
data
on restricted,narrowrangeof exploited experimental (Munday, 1963,Yao, 1974).
Othersusedthe dependenceof turbulencefactor on flame and vent outflow Reynolds
in
number orderto take into accountthe changesof P during explosion (Chippett,1984,
Molkov, Dobashi,SuzukiandHirano, 2000).

The disadvantageof using this approachis the difficulty to estimatethe averagevalues

of P from suchresults.FurthermoreZalosh (Zalosh, 1995) statedthat there is no clear
correlationbetweenP= constantor P=f (Re) approachesobtainedby different authors,
The authorelaboratedthat the assumptionsmadein the formulation of detailedmodels
concern:assumptionaboutthe compositionof the ventedgas,assumedflame geometry,
the empirical parametersandcorrelationto accountfor turbulenceenhancedcombustion
 66

and flame acceleration induced by flame instabilities. Further, this method requires
users to determine turbulent Bradley number, Bt and thermodynamic (1,,, Ej, qj) as well

asthermokinetic(&j) databeforeapplyingthe formulabelow:

p -UA

)OA 1+0.5. A.c,,,

A.(] + 1,
Br,.V3--6; (Ej -1)
Vy-1(0.73Smj).
r,.V21'
(3.22)
63
a(l+IOVY)0.4 (0.73.S,, (Ej - 1)
cu . . j).

As far as the author's concern,there are basically two different approachesof the
explosion venting flow modelling applied. In order to obtain the mathematical
expressionsfor flame flow and pressuredevelopmentduring venting explosions,it is
necessaryto establishthe appropriatethermodynamicand physics fundamentalsi.e.
conservationof mass,gas law, energybalanceetc. For a vent to give no increasein
overpressureother than that due to the pressuredifferencecreatedby the massflow of
unbumt gasesthrough the vent, the vent massflow rate is assumedto be equal to the
maximummassburning rate of the flame and this considerationshould be usedas the
design massflow through the vent. This approachin calculating the maximum vent
flow could be said to overestimatethe maximum flow as the assumption of the
maximumflameareais very conservative.

Experimentsmadein the laboratoryshow equationsfor the gas flow through an orifice

apply equallywell to flow througha vent (Nagy andVerakis, 1983).Sincethe vent can
be treatedas an orifice, the incompressibleflow equationthrough the orifice can be
applied but somehow,the assumptionof using this equationis only valid Up to Pred
200 mbarthenthe full compressibleflow orifice shouldbe used.

This approachwill be explainedin two different methods.By Method 1, it is assumed

that the maximum overpressureoccurswhen the ventedunburnt gas flow rate is at a
maximum and this equal to the maximum massrate of burning at the flame front, Mb
which has beenapplied by Swift (Swift, 1983),Munday (Munday, 1963) and Molkov
(Molkov, 1995)beforethe modificationmadelater in 2000. The equationis as follows:

Mb
= SoA, p,, (3.23)
 67

2
where S,,is burning velocity (m/s), A, is the internal surface area of the enc osure rn,
p,, is the (kg/m 3) The assumption is valid as the flame
and unburnt gas velocity .
the
approaches wall at higher Predas the flame area is closer to the wall. The maximum

vent mass flow and pressure loss always occurs before the flame exits the vent at which
during the period when the flame is expanding inside the vessel and pushing unburnt

gases ahead of the flame due to the thermal expansion of the burnt gases. Once the
flame exits the vent, the burnt gasesare free to escapeand expand outside the vent
the
whereas expansion inside the vesselceasesand hencethe flame speedsinside the
vesselwill fall to S,,.Further,the burnt gasflow throughthe vent is at a low gasdensity

and thus, will lower the pressureloss at the vent entrance.Thus, the prediction of the
maximum ventingoverpressureis reducedto that of predictingthe maximum unburnt
gasflow throughthe vent and its associatedflow pressureloss.

However,Method 2 assumedthat the maximum overpressure

occurs when the vented
unbumt gasflow rate, mbis at a maximum and this equalto the maximum unbumt gas
displacedflow by the flame front, S,. This assumptionhasbeenapplied for correlation
developedby Runes(Runes,1972)and Bradley andMitheson(Bradley andMitcheson,
1978a).The equationgiven is slightly different from Eq. 3.23 as is about 6.5 times the
massflow of the first method as it takes the effect of (E-1) where E is the expansion
ratio. As S. is closeto flame speed,Sr in value, this approachis only slightly lower in

massflow ratethanif it is basedon flame speedsvalue.

J'nb ý- SgAs
pa :--Su(E 1)A,
p,, (3.24)
-

The approachof assumingthe maximum flame area, Af to be A, as one of the

characteristicin defining maximum burning rate maybecan leadto the overestimateof
In sphericalvessel,the flame would only propagatesabouthalf of the
the overpressure.
diameterat P, = 100mbar and hencethe surfaceareais abouta quarter of A,. Taking A,
to calculate how much bigger the flame would take when venting occurs has been
initiated by Runes(Runes, 1972). For a safe design of a vent, Runes (Runes, 1972)
suggestedthat the important parameter needed to be considered is the required
maximum burning rate and this can be given by the maximumflame surfacearea, Ar
and by this assumption,Runessuggestedthat this could not be larger than the internal
surfaceareaof the enclosures,A,.
 68

For sphericalflame propagationat centralignition, the massburning rate of the flame,

Mb equal to the incompressibleflow equationthrough the orifice would give using
Method 1 andMethod 2 respectively;

Mb = A, S,,p,, = CdeA,(2p,, P lf-' (3.25)

..

Cd'FAv (3.26)
Mb= A, S, p,, = A, S,,(E - I)p,, ý-- (2p,,P,,,
d)0'5

whereCdis dischargecoefficientand Eis compressibilityfactor. It can be arguedon the

coefficient discharge,Cd value as the Cd valueis alwaysregardedas a constantof 0.61
for a sharp edge orifice but this would not apply for lower K value where Cd can
increaseup to 0.7 or more.In theory, Cd is dependenton the designof the vent and K.
However,this doesnot changethe fundamentalcharacteristicsof the flow, as in orifice
plate flow meteringcompressibilityis treatedby addingan expansibility factor, s, to the
basicincompressibleorifice flow equation,as shownin Eq. 3.25and 3.26. The value of
e decreasesas Predincreasesand is 1.0 for incompressibleflow and about 0.8 just prior
to sonicflow at the nozzle.The compressibilityequationis givenby BS 1042as:

1P a "d
I- (0.41+ 0.35 2) (3.27)
+ P,,,
K, 7(pi
d

wherey is a specificgasconstantand Pi is the initial pressure(bar). Another parameter

that alwaysassumedto be constantis the unburntdensitywherethe value is takenas 1.2
kg/M3 This is unjustifiablesimplification as the density is the linear function Of Predin
.
absolutepressuretermsfrom the gaslaw. Thus it shouldbe written as:

p
redMW (3.28)
RT

where MW is the molecular weight of air (0.029 kg mol"), R is the gas constant
(82.0552 x 10-6M3bar/Kmol) and T is the temperature(K). The result of taking the
densityas a functionof Pedis to increasethe massburning rateand thus requirea larger
vent areafor the sameoverpressureor a large overpressurefor thesamevent area.
 69

Again, the problem arises as temperature is also assumed constant. During the initial

portion of the explosion, the vented gases are cool. As the flame expands, the gases
flowing through the vent will soon have the high temperature of the combustion

reaction and basedon the gas law, high temperatureassociatedwith high pressure
attained.The assumptionof constanttemperatureis not valid but the computationof the
temperaturerise due to compressionis difficult without constantvolume combustion.
However, Lunn (Lunn, Crowhurst and Hey, 1988) has introduced approximate
temperatureasa function of P,ýd.
empiricalmethodto estimatethe compressed

r. 286
)/P
T= 298[(P,,,,+ Pi i (3.29)

However,the rise in temperaturedueto the pressurerise is smallerthan in pressureand

by using a relatively high T of 300K, the error will be small and conservativeas the
density will be overestimatedbasedonly on Pred-SubstitutingEq. 3.29 to Eq. 3.25 and
Eq. 3.26 would give

( ) 1/2
A, 0.00243S, P,, Po
= d+ (3.30)
P,
A, CCd
-,d

( Pred
A, 0.00243S,,(E - 1) + PO
= (3.31)
A, -cCd
Pred

Noted that A, and V2/3is in relation in terms of A, =C jjV3 The constant value is given
.
in Table 3.2 below. For spherical vessel, A, = 4.84 V2/3and this will result to

0-0 118Su 112

1 Pred +A
P,,
(3.32)
d

( P,, ) 1/2
0.0118S,,(E - 1) d
+ PO
(3.33)
K -ECd
Pld

Equation 3.32 and 3.33 are suitable only when the flow through the
vent is subsonic.
When Pred > 1.89 Po, the flow is chocked and sonic flow is
occurred. For spherical
vessel it gives,
 70

S,, P,.,, + po

dz
0.0145 1/2 (3.34)
KT

1)( Pred + PO
1 S,,(E -
0.014.5 1/2 ý
P,,
(3.35)
KT d

Howevertheseapproachesdo not predict the influenceof P, and turbulencefactor, 0

can be introducedin order to simulatethe effect. For high P, the breaking of the vent
will causethe pressurewavesand generatedturbulencewherethe net effect from these
actionsis increasingthe rate of burningat the flame front. In computer modelling, the
turbulencefactor canbe usedto give an effectiveturbulencevelocity, St = PS,

It seemsthat the underlying theory behind venting explosion makes other published
correlationsgiven by the previous investigatorscan be compared.It is important to
comparethe theory approachdiscussedabovewith the correlationsoffered by the vent
designstandardsi.e. NFPA 68 andEuropeanGasVenting Guidance2006.

It is instructiveto concludethat the reactivityterm usedby the vent designcorrelationis

the burningvelocity, S,,insteadof the KG values.In reality, the unburnt gasesinside the
vesselarenot ventedout at the sametime when the burnt gasesstartedto eject through
the vent wherethe unbumt gasesleft and trappedat the comer region of the vesselin
the caseof non-sphericalvessel.Oncethe burnt gasesare vented, there is no further
internal expansionof the flame and the flame speedsshould be reduced to laminar
burning velocity, S,,.At this state,the pressurewill be associatedwith the subsequent
combustion of the unbumt gases at & and the assumption for the maximum
overpressureoccurredat the maximumvent unburnt gas flow rate is valid. Below are
the studied correlations along with the proposed equationsto be compared to the
publishedexperimentaldatathat will be discussedlater.
 71

Table3.1 Studiedcorrelationsfor ventingof gasexplosions

Correlation Reference
1 (Bartknecht,
[ 0.1265log, K, 0.0567 0.175(Pstat 0.1) Bartknecht
- + - V
p 82 p 0,1572 1993)
IlýU

CA Swift (Swift, 1983)

A, -'
0.5
Pmd

P,
-,d=
for P., k1 barg
2.43(A/ý)-'0*69" Bradley and Mitcheson
(Bradley and Mitcheson,
P,. =12.46(A/ý)-' forP: gl barg
d 1978a, Bradley and
Mitcheson,1978b)
; r,,
d = Br, -2" for <1 Molkov (Molkov, 1995)
'T,, d

0*5 for
; rr,
d= 7- 6Br, 'Tred
>1

1 Pmd + PO ) 1/2 Method I

0.00243CSI,
Pred
K

I ( Pred
+ PO
1/2
Method2
0.00243CS,(E - 1)
P,,
K ECd
d

3.4 Comparison of Bartknecht and Swift correlations as been adopted in

NFPA 68 using derived methodson the samebasis.

To justify the applicabilityandvalidity of the proposedequations,it is useful to directly

comparethe derivedmethodswith the correlationsgiven in NFPA 68 i.e. Bartknecht's

equationandSwift's equation.From Eq.3.7, Bartknecht'sequationcan be expressedas,

I=C
-0.5817
(3.36)
K 3Pred

whereC3= 0.164for methane,0.200for propane,0.212for town's gasand0.290for

hydrogen.
It shouldbenotedthatthereis a relativelysmalldependency
of K on mixture
reactivity which is most unusual basedon the fact that hydrogenis the high burning
velocity mixture comparedto methaneand propane.For Method I and Method 2, Eq.
3.30and 3.31 canbe transformedin term of the constantvalueterm as,
 72

1=
Cl C2 E-1 Pr ed -"" (3.37)
K

0.5 /(C 0.5 /(C

where C, = (p for Method I and C, = (p (E-1) for
d2o-")S,, d2o-5)S,,
Method 2. C2 = 4.84 for sphere, 6 for cube, 6.3 for rectangular, 5.81 for cylinder with
L/D =2 and 5.54 for L/D=l Ci is the same derivation for Swift's C constant. If Su= 0.4
-
m/s for methane,0.45 m/s for propane and 3.1 m/s for hydrogen, Cd = 0.61 and p=1.2
kg/M3, it would give C, = 0.0016 for methane, 0.0018 for propane and 0.012 for
hydrogen respectively for Method 1. For Method 2, with the same value of %, Cd and p
for appropriategases,C, = 0.0104 for methane with E=7.5,0.0126 for propane with E

= 7.98 and 0.078 for hydrogen with E=7.29. If these C1 values were compared to
Swift's C constant and C3 in Bartknecht's equation, it will give the value as shown in
Table3.2.

Table3.2 Comparisonconstantvalue derivedfor Method I and 2 with respectto Swift's

C, fuel characteristicconstant
C3
Fuel C1 CI C
Method I Method 2 Swift's Bartknecht's

prediction prediction constant equation

Methane 0.0016 0.0104 0.037 0.164
Propane 0.0018 0.0126 0.045 0.200
Hydrogen 0.011 0.068 Not available 0.290

Notedthat it shouldbe includedC2for geometry'seffect andfor cube,it would give the

combinedconstantCIC2= 0.0097,0.011 and 0.114for methane,propaneand hydrogen
respectivelyfor Method 1. Method 2 would give CIC2= 0.062 for 0.071
methane, for

propane and 0.41 for hydrogen. To get agreementfor the turbulent effect or self
is
factor,0,
acceleration the ratio Of C3/CIC2
as Bartknecht'swork usedin
calculated
larger vesseland it was suspectedthat self-accelerationis already presentduring the
explosions.Thelist of predictedP is given in Table 3.3 below.
 73

Table 3.3 List of predictedP with respectwith Bartknecht'sC3constant

Fuel CIC2 CIC2 C3 P P

Method 1 Method 2 Bartknecht Method 1 Method 2

Cuhe
Methane 0.0097 0.062 0.164 16.9 2.6
Propane 0.011 0.071 0.200 18.2 2.8
Hydrogen 0.114 0.41 0.290 2.5 0.7

Sphere
Methane 0.0078 0.050 0.164 21.0 3.3

Propane 0.0087 0.0569 0.200 23.0 3.5

Hydrogen 0.068 0.416 0.290 4.3 0.7

Rectangular
Methane 0.01 0.066 0.164 16.4 2.5
Propane 0.011 0.08 0.200 18.2 2.5
Hydrogen 0.078 0.49 0.290 3.7 0.6

Cylinder for L/D =2

Methane 0.0093 0.061 0.164 17.6 2.7
Propane 0.010 0.073 0.200 20.0 2.7
Hydrogen 0.072 0.45 0.290 4.0 0.6

From the listed P value for both Method I and 2, it seemsthat Method 2 gives a good

reasonableagreementfor turbulent factor for methaneandpropaneas the valuescloser

to the turbulentfactor given by other works (Munday, 1963,Pasman,Groothuizenand
Gooijer, 1974,Yao, 1974)but not for hydrogen.HydrogenusingMethod I andMethod
2 give over prediction in respect to Bartknecht's results and these are unusual
experimentalresult.It should be noted that Swifts equationgives P=7 (Swift, 1983)
for turbulentvalueduringexplosionandthepredictedP valuesin Table3.3 werelower
than Swift's turbulentvalue,TheseP valueswill be comparedwith the best-fit P values
from the tabulatedexperimentaldatalater.
 74

3.5 Reviewof the published experimental data.

A significant amount of data concerning vented explosion of methane-, propane-,

hydrogen-, acetone-,ethylene-andtown's gas-air mixtures were found in published
literature. The reported470 experimentswere performed in a wide range of initial
conditions: enclosureswith various volumes (0.12 - 200 3)
in and shapes (cube,
rectangular,cylinder and spherical)at variousbursting vent pressure,Pgat(O- 500barg)
with different length-to-diameterratio (L/D=1-4) and various location of ignition
sourcesresultingscatteredrangeof maximumreducedpressure,Pred(0.014-33.7barg).
The data near or at stoichiometricmixtures with air were studied.Experiments with
centralpoint ignition, point ignition near the vent and at the rear wall, as well as with
plane and jet ignition, were processed.However, as statedin ATEX and European
Guidelinesfor venting,it is said that centralignition producedworst casescenarioand
due to this, only experimentaldata at centralignition will be investigated.Someof the
publishedresultsdo not havesufficient internal details to calculatethe internal surface
area,A, and thus,usingV2/3:zýA, is appliedto dataif the dimensionsare not available.
For those without detailedinformation on the geometry,an assumptionof LID =2 is
used for the calculation purposes.The list of constant used for the correlations'
calculation presentedin Table 3.4. Kr, for calculation is excerptedfrom NFPA 68
(1978): Ko values for gasesdeterminedwith IOJ electric spark. The details of all
experimentaldataaregivenin Appendix A (TableA. 1).

Due to lack of agreementsof the publishedcorrelationsto be applied on the venting

design purposes,the collected published experimentaldata will be comparedto the

main equations:Bartknechtand Swifts equationas beenofferedin NFPA 68, Bradley

andMitchesonasthe equationoffers for the safeventing approachand Molkov as been
recommendedto be oneof the alternativesto be usedfor the ventingdesignin NFPA 68
(2002). For other equations,it will be reviewed basedon the necessitYof the data
comparisonwhen situationpermits.All graphswere plotted by PredvK with respected
correlations mentionedin Table 3.1. The data will be categorizedaccording to the
fuel/air type, the vesselshapeand volume andPvis between0 to 500 mbar asthis is the
validity in using Bartknecht'sequation.The reasonto follow closely the Bartknecht's
range of limitation/validity is due to the wide range of vesselvolume to be applied
 75

which V is up to 1000 m3 where others are not. In case of P, > 500 mbar, the different
analysis will be formed in order to predict the suitability, availability and validity of
thoseequationson the experimentaldata studied.All data sets are distinguishedwith
shapeof the vesseland different volumeof the samevesselshapei. e. ci for cubic, o for
sphere,0 for rectangularand A for cylinder. This is done to show whether there are
independentvolume effects which is not been accountedin K and if there is any
differencesdueto the vessel'sgeometry,for the sameK andvolume

Table 3.4 KG value used for vent area calculation. All data were excerpted from Bradley

and Mitheson (Bradley and Mitcheson, 1978b) except for acetone-airdata (Molkov,
Dobashi,SuzukiandHirano,2000)
Fuel-air n-dxtures KG S, C P. YU0 7 bo
(bar-m/s) ( M/S) ( bar)1/2 A

9.5 % C114-air 55 0.43 0.037 7.52 1.38 1.18

4.0 % C3H8-air 100 0.45 0.048 7.98 1.365 1.25

29.7 % H2-air 550 2.70 - 7.29 1.40 1.25

6.53 % C2114-air 243 0.68 0.048 8.06
25% Town gas-air 150 1.22 6.64 1.38 1.18
-
Acetone-air 84 0.39 0.048 7.96 1.36 1.25

The relationshipbetweenA, to V213

asA, =C V2f3is shownas follows:
Shapeof the volume Constant,C
Sphere 4.84
Cubic 6
Rectangular 4n +2 wheren= LJD
Cylinder 5.81 for UD =2
Cylinder 5.54 for LID =I

Further, since the applicability of Swift's equationis UPto Pred 100 mbar which is
-
subsonicflow regime,the sonic flow will be appliedfor Ped> 900 mbar. The sonic flow
applied when the pressureupstreamof the vent reacheda critical pressure,Pc which
 76

equalto 900 mbarandthe velocity of the gasflowing throughthe vent becomessonic.

The flow equationfor sonic flow changesdirect proportionalbetweenthe massflow
and the upstreampressure, On
Pred- the Pred
vK graphs, in the case of Swift's equation,
the Swift line with Pred" exponent is plotted UP to Pred< 900 mbar and then the line

with Pred1*0exponentis plotted for sonic flow.

To the author'sknowledge,a quite recentcomprehensivecomparisonon the published

correlationon ventingdesignhas beenmadeby Razusand Krause(Razusand Krause,
2001) From their analysis,on the experimentaldata and by using cylindrical vessel
.
with central ignition as fictitious examplesof vented deflagration, they stated that
NFPA 68 (NFPA68,2002) and Bradley and Mitcheson correlations (Bradley and
Mitcheson, 1978a,Bradley and Mitcheson,1978b)gavehighestPredand suitable for a
worst caseconsiderationonly. In a casewherethe static burst pressure,P, is involved,
equations given by Bradley and Mitcheson (Bradley and Mitcheson, 1978b) and
Molkov (Molkov, 2001) seems to agree more on the experimental data for
hydrocarbon/air yet the hydrogen/air needed more comprehensive experimental
examination.Computationalmodellingon ventingwould not be consideredin this case.

3.5.1 Comparison of Bartknecht and Swift's correlations with published

experimental data.

To get more in-depth investigation on the discrepancyon both equationsoffered in

NFPA 68, the published experimentalvented explosion data for methane,propane,
hydrogenand town's gas were shown in Fig. 3.4-3.7 with Bartknecht's and Swift's
correlation.All data is been correctedto P,, = 100 mbar to allow the minimum P, in
Bartknecht'scorrelationto be appliedThe data points indicate the shapeof thq
explosionvesselanddifferentvolumesof thesamevesselshapeasmentionedabovei.e.
[3for cubic.This is doneto showwhetherthereareindependentvolumeeffectswhich
not correlatedby K andwhetherthereareanydifferencesdueto vesselshape,for the
sameV and K. The datasourcesare too numerousto be listed and not specifically
identifiedin the graphs,as the objectwasto concentrate
on vesselshapeandvolume.
Figure3.4 and3.5 showthat mostof thedatafor spheresandcubic vesselsaresimilar
 77

and scatteredaroundBartknecht'scorrelationline. As mentionedearlier, Bartknecht

developedhis correlationbasedon the experimentaldata obtainedfrom 10 rn3 cubic
vesselvolumeof propane/air.However,asshownin Fig. 3.5, someof Bartknecht'sdata
of propane/airat V= 60 m3are well below the correlationline which is not a general
trend for cubic vessel,For hydrogen,only cubic vesselsfrom Bartknecht'sdata were
processedsince other works did not have the same hydrogen concentration as
Bartknechti.e. 40 % v/v.

10 m v. 1m*
V. Ims
,,_*ýV. 22m3
0
ED V- 30m'
-- -- Bartknecht
rj - Swift
- Sonic flow line

AO 13

Cz A Sonicventing
.0 ........ .................... ................................................
Subsonicventing
CL
13

Methane -air

10 100
K
Figure3.4Methane/airexperimentaldataasa function of K.

As beenillustrated in Fig. 3.7, the rectangularvesselsat lower K and subsonicflow

gave poor fitting for both equations,suggestingthat there is an influence of vessel's
geometry in the developmentof venting gas explosion, The use of V 2n in vent
coefficient probably not the best approach to representthe influence of vessel's
geometryin venting gas explosioninsteadof total surfaceareaof the vessel,As. Later,
the discussionbetweenthe use of V2/3 and A, will be discussedin details in order to
determinethe bestapproachto characterisethe influenceof vessel'sgeometry.
 78

00 V-1m,
13 V- 20
ED (D V-10m,
10 30M3
00V .
60M3
E3 E) V-
Bartknecht
Swift
..........Sonic flow line
Ea
(D O(D M0
(1) CO
00
ie
C13 Sonicventing
.......................... =din .... m..................................................

a.
/
//te
/'
Subsonicventing

/
/', e

Propane-air
0.1
10 100
K

Figure3.5Propane/airexperimentaldataasa function of K.
V=1m 3
10
Bartknecht
Swift
Sonic flow line

Sonicventing
C51 - I
....................................................
.0:..................................
*? Subsonicventing
CL

Hydrogen-air
10 100
K

Figure3.6 Hydrogen/airexperimentaldataas a function of K

 79

10 * V. Irr?
V-1fe
14 Vc loff?
Badknechl

21 Sonicventing
ca I-
..............................
.. .........................
LP Subsonicventing

Towngas-air
10 100
K
Figure 3.7Town's gasexperimentaldataasa function of K

Lacking of dataat the sameconcentrationas Bartknecht'swork is the reasonwhy only

four data presentedin Fig. 3.6. There are -16 data points with higher Predthan
Bartknecht'scorrelationwhich wasonly fitted to his own dataset and includedno other
publisheddata.The results show that Swift correlationhas -5 data points with higher
Predthan predictedand hencewould be a safer prediction than Bartknecht, thus it is
recommendedto be usedand the limits on its applicability are also removed,providing
a P, term is added.

It is the main objectiveto compareall experimentaldatawith the publishedcorrelations

to showtheir validity and applicability.Further,the proposedtheoriesbasedon Method
I and Method 2 on venting will also be plotted on the graphsin order to make the
comparisonbetweenthe publishedcorrelationsand how the equationsfitted well with
the published experimental data. It should be noted that there is no turbulence
factor, P be addedto the proposedequations(Method I and Method 2) in
enhancement
order to show at what extent the turbulence generated during the explosions.
Afterwards,theturbulentenhancement
factor, 0 will be addedto the proposedequations
that best fitted with most of the data.It is statedthat Swift usedP=7 and Bradley and
Mitcheson appliedP=5 to representthe turbulencegeneratedduring the explosion in
order for correlationbe well agreedwith experimentaldata.
 80

3.5.2 The influence of turbulent enhancementfactor, 0

At this stage,turbulencefactor, P is not taken into account. As been mentioned

previously,the level of turbulenceinducedin the ventedexplosionhasto be determined
experimentallyasthe determinationof turbulentfactor hasnot yet agreedto most of the
practical application (Chippett, 1984, Molkov, Korolchenko and Alexandrov, 1997,
Munday, 1963,Pasman,Groothuizenand Gooijer, 1974,Yao, 1974).The disadvantage
of usingthe suggestedapproaches is the difficulty in estimatingthe averagevaluesof D
from suchresults.When the vent areais low, it is not sufficient to vent the unburnt gas
out from the vessel.The flame will be distorted,increasingthe flame area.Thus, this
criterion will increasethe rate of pressurerise and hence,the pressurewill build up
rapidly. Decreasingthe vent area, i.e. increasingthe vent coefficient would initially
increasethe maximumpressureas the flame becomesunstable.During the onsetof the
venting, the flame will first start a Helmholtz oscillation; the pocket of burned gas
within the vesselundergoesbulk motion towards and away from the vent opening.
Consequencefrom these oscillations, the rate of volume production by combustion
increasessufficiently for the internal pressureto increase,the oscillations themselves
being gradually damped out as the flame expands. In this trend, turbulence was
generatedin the shearlayer betweenthe out flowing burned gasand the unburnedgas
within the vessel.The Helmholtz oscillationsthen induced Taylor instabilities. Taylor
instabilitiesweremore profoundin larger vesselas discussedby Solberget al (Solberg,
Pappasand Skramstad,1980).

From Fig.3.8 and 3.9, turbulent enhancementfactor, 0 =1 and Cd = 0.61 are used in
order to correlate the same Cd constantusing by other comparativeequations.For

Molkov's equation, is treatedthe sameway sincethe equationis derivedbasedon its

own correlationas the constantparametersarenot specified.It is illustratedthat the

turbulencefactor is neededcompulsorilyto be addedinto the proposedequationin
orderto takeinto accountthe effectof staticburstingpressure,
P, of the vent andthe
duringtheventingexplosions.
existedof theturbulence
 81

CH4/adr
10

A
* V--'D. 8 rTO(DragosaNAC)

x V_-22 m3(P. F. Iborrle)

00 V--2&6 m3(BurJdand)
V '30 m3(Bartknecht)
V-_34ni3( RGLZalosh)
x46 Bartknecht
SWft
----
---- Bradley & Mtctxrcn

....... Molkov
'0
Method 1
40 x
Method 2
0.1
1 10 100 1000

Figure 3.8 Methane/air at various geometries without

lo C3HWair 13 \/=l 0 m3(Bardqxx*ft)

0 Wl 0 m3(Donad)

N W22 m3(RF. Thome)

0 \k. 25 m3(8romma)

z V-10 m3(Barfinedt)

0 WM m3(Dork-d)

iI/X, # \k-35 m3(Solberg)

0 \A.40 m3ftomma)
V660 m3(Bardoectl)
x+
V660 m3(Donat)
f0
/" *1 V-70
0 0 m3(Bromma)
+0* V68lm3(Howards&KArabis)
00
\ýL-" m3ftomma)
+00
0 V--)M. 8 m3(P. F.Thome)
0:
4 Bartimedi

---- BradleA MtheSon

....... rvblkov
0.1
- MAIOd 1
0.1 1 10 100 off
Mtiod 2
Ka

Figure 3.9 Propane-airin various geometrieswithout

From those figures above, it is clearly seen that Bradley and Mitcheson's equation has a

good agreementwith the experimentaldata comparedto other studiedequations,except

for V= 34 m3at sonic venting with low K for methane/air.The implication of this is the

adequate and safe protection in venting design can be applied for Bradley and
 82

Mitcheson'sequationbut not in economicand practical application.It is crucial to

develop the venting design equation that can provide safe, adequate and economical

valuesin practicalapplication.

For smoothly opening vents, Yao (Yao, 1974) recommendedto use 0=3 and for
bursting diaphragms,P=4 while Pasmanet al (Pasman,Groothuizen and Gooijer,
1974)gaveP =3 for small vent area( Av :50.1 rn2) and P =1 for the larger openings(A
> 0.1 2).
m Swift and Epstein (Swift and Epstein,1987,Swift, 1983)and Swift (Swift,
1983)indicatedthat, exceptfor the mostextremecases,a turbulentenhancementfactor
of 5 is sufficient to envelop the effect of vent opening pressureand used to give
deflagrationcharacteristicvalue, C in calculatingthe desiredvent areaof the protected
Further, Bradley and Mitcheson (Bradley and Mitcheson,
equipments/containments.
1978a,BradleyandMitcheson, 1978b)alsoindicatedthat in orderto get the theory and
experimentto be in agreement,a turbulent factor of 5 would have to be assumedas
beingcreatedby the vent. This recommendation applieson the initially coveredvent in
their correlation,known assafevent designfor ventedexplosion.

If turbulenceoccursin the explosion or appearsin the beginning,it can be taken into

accountusing an experimentallyderivedturbulenceenhancementfactor, P so that the
reactivity is PSu.Turbulentis generatedasthe unburntgasis pushedaheadof the flame
front when the vent is openedwhich can give rise to the additional back pressureor
driven past the obstaclesor over surfaces.The effect of turbulenceis well known and
been investigatedby various past researchers(Burgoyne and Wilson, 1960, Solberg,
Pappasand Skramstad,1980,Andrews,BradleyandLwakabama,1975,Canu,Rota and
Carra,1990).Turbulencecreatedby the ventopeningor the obstaclescan be includedin
Eq.3.30 and3.31 andbecomes;

( ) 112
1=0.00243CPS,, P,, Po
d+ (3.38)
K ECd P,,
d

I=0.00243CgS,, 1/2
(E 1) Pred + PO
(3.39)
K -Cd
 83

Figure 3.10,3.11 and 3.12 showed the exampleson how the possible P had been
processedin order to fit with the experimentaldata. The summary of the average
turbulentfactor,P for different vessel'sgeometryandfuels given by Table 3.5,

Table3.5 Summaryon averageturbulentfactor,P for different vessel'sshapeandfuels

Vessel's Methane Propane Hydrogen Town's Acetone Ethylene

shape gas
P P P P P 0 P P P P P P
MI M2 ml M2 MI M2 MI M2 M1 M2 MI M2
Cubic 20 4 20 3 -20 5 - - - -
Sphere - - -20 4 - - 20 3 - -
Rectangular 20 4 3 >20 1 10 1 10 2 I- 7 2
-20 -
Cylinder 20 3 15 13 1- 1- 1- 1- 1-
j

The turbulentfactor, 0 tabulatedin Table 3.5 seemsto agreewith most of the previous
but slightly lower that beengivenby Swift (Swift, 1983)andBradley and
experimenters
Mitcheson(Bradley and Mitcheson, 1978a,Bradley and Mitcheson, 1978b).However,
if the valuesin Table 3.5 is comparedto valuescalculatedusing Bartknecht's constant
in Table 3.3, it can be said that both values seemsto have a good agreement.This
implies that the vesselvolumeis the importantrole in determiningthe Predin ventedgas

explosionsassignificant self-accelerationis inducedin the larger vesselbut having less

effect in smaller vessel as been depicted by McCann et al (McCann, Thomas and
Edwards,1985) It shouldbe noted that the P, effect is not countedin the Method I
.
and Method 2 approachesand it is believedthat the turbulent factor can be used to
representthe effect of flame propagationby the vent flow dueto the burstingvent.
 84

V-0.2 nL3 CH./air -Ml * V-02 m3 CHIair -NV

*
10 V-0.4 n13
lo 0 V-0.4 nl,3
0
V-0.6 m3 * \A-0.6M3
0 A. ,
V-2.8 m3 o \062-8m3
0
V-1 1.2 m3 0 V-1 12 m3
0
V-10-8 m3 0 V-0.8 ff%3
0
\A-26.6M3 V6Q8.6m3
Tud)Jea of 2 o
TurbLAOMof 20 0
0 TurbilaV of 3
TurbulemOf10
to
...... . TuWent of 7 00 ..... .. TurbLAom of 4 o
TurbLJM of 5 0
-TLrbLAentof5 to 0'
8 /, 8 $00,
8 1*0 0
0.1 - , 0.1
0.1 1 10 100 0.1 10 100
K

Figure 3.10 Methane/air with turbulent factor, 0

lo Tomygas
KM 10 T--19-

V-0,23 niG
0V-0.23 M3
am
.01 V-1.0 rTo
0V-1.0 rra
V-1.47 rTG a
V-1.47 ffl3
V- 28nG am
0V-28 n13
TLYOAM at 2
of 20
TL#tLAerit
0.1 4t
---- TurbUýt of 3
TabLdemd 10

TubAM d7 ... ... TurbLM ot 4

.......
TurbLAcro
of 5
TjrbLAertof 5
OL01
0 10 IOD 0.01
0.0 (11 1.0 1010
K

Figure 3.11 Town gas/air with turbulent factor, P

CF6/air 10 C3Wair
10 mi
a V. 077 ,
I NE a
V-a77
.,, a
V-1.0 V-1.0
cr. , ..0
x V--20 x V-20
x x++ Ix
I ool x V-10LO V. lao
/ + V-2ZO V-220
a x a/ -x
V. Oao , V-MO

+ /+ a
+ V-60LO W8M
/, ' o
0.1 RrOLdert d 2D Tb, J«t of 2
+,, /, /, o
TLMLMM 09 3

+1
---- TLgbLA&ld 10 ----
0 TLIrtLAertCt 7 aa TLrbLffl of 4
I
1- .
.
. TurbJert d5 / " TUrtLiErt Gf 5
0.01 - 0.01
0.1 1 10 100 a1 1 10 100

Figure 3.12 Propane/air with turbulent factor, 0

However, it seemsthat Method 1 with high 0 do not give satisfactoryresult to fit the
experimental data. In this it
case, can be argued on the application of S,, (E-1) and Su
terms for demonstratingthe unbumt gasaheadof the flame characteristic.This different
approachto the maximum vent flow rate of unburnt gasesis a crucial in vent prediction
assumptionas the difference is betweenSg Sj- S,,
or and where usually a factor of 7 or 8
for hydrocarbonand hence(E-1) is - 6.5. As S,,(E-1) is close to Sf in value, Method 2
 85

will give slightly lower vent massflow ratewhereit is clear that basedon flame speeds,
Sf. The correctionfactorsfor expansibility and density as a function of vent areaand
Predshownthatthe correlationfitted well with the datain the regioncomparedto others.

Molkov (Molkov,Dobashi,Suzuki and Hirano, 2000, Molkov, 1995, Molkov, 2001,

Eber,Tamanini and
Molkov,Korolchenko and Alexandrov, 1997, Molkov, Grigorash,
Dobashi,2004) in his universalvent sizing correlation employedthe inverseproblem
to
method correlate the turbulencefactor, X and generalised dischargecoefficient, ýL.
the level of turbulenceshould grow with flame scaleand
From author's understanding,
vent area. Further, the maximum explosion overpressureshould correlate with the

squareof the deflagration-outflowinteraction number, (Z/,U)2 Their work suggested

.
that the deflagration-outflow-interaction
number is dependantof fractal theory and with
that criterion, Reynolds number is no longer needed in vented explosion
correlations(Molkov, Dobashi,SuzukiandHirano,2000).

3.5.3 The influence of the vessel'sgeometry

Both explosion parameters are dependent on the design of the vessel. In general, the

values Of Predand dP/dt are changed because of alterations in the amount of heat loss

from the flame to the vessel walls. The geometrical term i. e. size/shape of the vessel is

generally written as the ratio of the vent area to some other representative area of the

system (the intemal surface area of the vented enclosure, A, or its volume to the 2/3

power). However, the vent coefficient, K method is the most often used for the

assessment of vented gas explosion and this parameter will be used in all graphs'

plotting. The size and shape of the enclosure influences the final pressure generated; for
instance higher pressures will be generated in long narrow enclosures than in more

nearly cubical ones but in practice, it is the characteristics of the explosion relief that

will determine whether or not a plant is damaged by an internal explosion (Cubbage and
Marshall, 1974). These two approaches in correlating the venting design do affect the

prediction result on venting compared to experimental data. Bartknecht's correlation

used the V2/3 terms to describe a characteristic of spherical or compact vessel
explosions, where the flame remains mostly in spherical shape during venting process

on the dependence of overpressure on the test vessel volume. If the spherical flame
 86

propagatesat a constantrate irrespectiveof the vesselvolume,there shouldbe no other

dependence Of Predon volumeotherthan K. Meanwhile,Swift's correlationis given as
Av to be a function of cross-sectional
areaof the enclosure,A, multiplied to reactivity
terms which includethe burning velocity, Suandturbulent enhancementfactor, P. The
useof the ventedvesselsurfaceareain vent correlationswas first introducedby Runes
(Runes, 1972) as the maximum possible flame area and was used to calculate the
maximumflamemassburningrate.

The useof A, seemsto be appropriatedto link with the vessel'sshapeas if the vessel
shapeis substantiallydifferent from cubic or when UD >-2, the flame touchesthe
wall before there has been a significant pressurerise and sphericalflame propagation
cannotbe applied to model the explosion.In practise,maximum pressureusually but
not always occurs when the flame contactsa wall of the vessel and the flame area
begins to decrease(Nagy and Verakis, 1983). In non spherical vesselswith central
ignition, Ellis (Ellis, 1928)showedexperimentallythat the shapeof the flame tendsto
assumethe shapeof the vessel.This observationis agreedby Nagy et al (Nagy and
Verakis, 1983)with cylindrical vessel.In a large sphericalvessel,the buoyancyof the
hot gasesmay distort the flame development.If the ignition source is not centrally
located,the flame front will contactthe nearwall before combustionis completeas the
flame is allowed to expand in one direction resulting in an elongated flame with
increasinglylargersurfaceareaandhence,fasterexpansionthan centrally ignited flame.
Consequently,the maximum rate of pressureand the maximum pressurewill be less
than when ignition is central becauseof greaterheat losses.For central ignition, the
flame will expandsymmetricallyin all directionsin sphericalshape.After sometime,
the flame beginsto distort when first flames contactingthe vessel wall and stretching
towardsthe ventbeforechangingits shapeto a cylindrical flame.

From the data collected,it is observedthose cubic and sphericalvesselsare the most

vessels'geometryusedin ventedexplosion.Assumingthe flame initially behavesas a

sphericalshape,it is observed theoreticallythat the initial burnt massare lessthan 10 %
,
for cubic, cylinder for UD =2 and worst casefor rectangularwhere
only 2.5 % of the
masshas burnt during the explosionprocess(in a casewhere P.,,, is assumedto be 8
bar) whenthe flame touchesthe wall, Sincethe massburnt rate is
proportion to rate of
pressurerise (dP/dt),it will reducethe flame burning areaand laminar burning velocity
 87

and thus, decreasesthe Pred.

At this point onwards, the heat losses from the burnt gases
to walls are larger which also distributed to reduce-Pred-Quenching of the name just

prior to reaching the vessel wall can be expected to have a significant effect on the KC!
value in very small vessels. The reduced flame area after contacting with the walls
influences the mass burnt rate which finally decreasing the amount of mass flow rate

through a vent. Increasing L/D ratio, i. e. smaller D for constant L, less mass burnt rate is

obtained which results to decreasePred.Thus in large LID vessel, the spherical flame
portion of the flame propagation occupies a negligible part of the explosion event. This
is why the standard cubical relationship involved in Kr, or K, (dust deflagration
t
constant) parameters only apply to compact or near spherical vessels. For spherical
vessel, 14% of initial mass or at overpressureof 1.12 bars for maximum pressure inside
closed vessel is assumed to be 8 bars, has burnt when the flames first touch the vessel
wall compared to about 7.5 % initial mass ( overpressure= 600 mbars) burnt for cubic
vessel. It was found that spherical vessels give higher Predrather than cubic vessel due
to a lower quenching effect to the vessel wall and since the mass burnt rate is proportion
to rate of pressure rise (dP/dt), it gives rise to an accelerating flame front and an
increasedP,,d. This scenario is shown in Fig. 3.13.

V-1 rro
10 - CF-Wair V=22 m3
x,
A- ýe V-Mrrß
0 Vml m3

- Bartkrech
/-0.
ty Bracley &
c]
#0

Mtcheson
All .... Wkov

Mathod 1

1a8 Mathod 2
li
O.j ý,,,,, -x, -,-
0.1 1 10 100 1000
K
Figure 3.13 Influence of vessel's shapeon methane/air explosion o Sphere and o cubic

From Fig. 3.13,it is shownthat the proposedequations(Method 1 and Method 2) gave

satisfiedcorrelationto fit most of the data when 0 is included. Even though Swifts
 98

equationregardedto be applied for Prd < 200 mbars, it shown that the application of the
equation can be extended. This is shown by most of the data fall under the Swift's line
as previously mentioned in Section 3.4. In rectangular and cylinder vessel, the majority
of the flame propagation is non-spherical as the flame is elongated to follow the vessel
shapewhere - 2.5% of the mass has been burnt when spherical flames reaches the wall
and hence, there are large heat losses subjective to convection heat. Since quenching
rate is higher, the reduced flame area influences the mass burnt rate which finally
decreasesthe amount of mass flow rate through a vent (Andrews, 2004). Thus for most

practical situations, there will be large heat losses from the explosion and venting will
occur after the flame has touched the vessel wall as shown in Fig. 3.14,3.15 and 3.16
on non-spherical vessels data scattered outside the published correlation lines even at
low K.

From Fig. 3.16, it was found that cylinder vesseldata gavebit scatteredplots for K> I
from the proposedequationsand published correlations.In a cylindrical vessel, the
flame shapeis divided into two periods.In the first period immediately after ignition,
the flameswerehemisphericand then half-spheroidalwith sectionsparallel to the wall.
In a very short period of time, nearly 80 % of the flame areavanishesbecauseof the

quenchingof wall parallelparts.As a consequence, the amountof expandingburnedgas

is suddenlyreduced,resulting a rapid decreasein heatreleaseaswell asP,,d(Starkeand
Roth, 1986).The reductionof the flame areaalso contributedto the low peak pressure
anddp/dtrnax.
Thus,for most practical situations,therewill be large heat lossesfrom the
explosionand venting will occur after the flame has touchedthe vesselwalls. Similar
results also obtained by Leeds work on a larger cylinder (Phylaktou, Andrews and
Herath,1990).Their work showedthat the point of deviation of the cylindrical vessel
from sphericalflame propagationis obtainedat the time the flame touchesthe wall,
causing lower dP/dt and peak pressure,It also concluded that the application of
sphericalflame propagationmodelling canonly be appliedto sphericalvesselif the full
flamepropagationis modelled(Phylaktou,AndrewsandHerath, 1990)
 89

10 V-4). 17 rml
0
0 V=O.18 n-LA
0 V60-5 n-L3

* V--35 nO
0 V=40 n LA
0 V--70 m3

* VL-81 ryLl
* V--2W n LI
0 W-203,8 rrB

- Ban knecht

....... SM ft

---- Bridey & Mtchaon

---- NbIkov

-- Wt hod I

0.1 6 MLAhod 2

OLI1 10 100 1000

Figure 3.14 The influence of rectangular vessels on P,, d for propane/air 0 rectangular.

10 Tavmgas/air

00,

'0'
13 13 V=0.23
V6-1.47
\162.8
dL
0 13 V= 1 M3
Barfimcht

0.1 00
Brachey& Mtd-eson
, '0 NUkav
----
/"--- Nbtx)d 1
6 M3tx)d 2
0.01
01 10 100
K
Figure 3.15 The influence of various geometries on Ped for town gas/air 0 rectangular, o

cubic, o sphere
 90

1,11Acetorelä r

V-0.19M3
Z0
A
V-OMM3
J
V- ZO rn3
Bm*nf)cht
-i ,
0/ , Smft
----
al Bradleyand
----
Mtcheson
.... Nbikav

Nbthod 1

Nbüiod 2

1 10 100 1000
K

Figure 3.16 Theinfluenceof the vessel'sshapeon Pred-A cylinder o sphere

3.5.4 ComparisonbetweenK and AýA, term for vessel'sgeometries

As been discussedabove on how the use of V25/A, (K) term is failed to give
satisfactoryresults in non-cubic and non-sphericalvessels,the term A, /A, will be
replaced the K term in order to investigatethe differencesin using both terms for
correlating the vessel's geometry in venting explosion. It should be noted that this
analysis is only involved the whole set experimentaldata of methane/airin various
geometries.

The data resultswill first be shown the P,,d vK for cubic, rectangularand cylinder
vesselsas illustratedin Fig. 3.17. Again, turbulent enhancementfactor, P is included
accordinglyfor all correlationlines. From Fig. 3.17,it canbe saidthat cubic vesselgave
satisfactoryresultsfor all correlationsbut not with rectangularvessel.The datascattered
aroundthe correlationlines, but mostly aroundMethod 2 and Bradley and Mitcheson's
correlation lines.This suggeststhat both Method 2 and Bradley and Mitchesoncan be
applied for safeventing guide in venting design purpose.However, the cylinder data
seemsto fit well below the correlationlines.
 91

clvýw
10 10
CLNC R
,c
0 WOý2-1 A/
0 W. 4-1

0 W16 nil
A
W2.8-3
V -I m3 \611.2 -1
V =22 m3 \ý20A M3
t.
x V---*) m3 w2b. 6 -1 *"
*
Bank-chi

....... Swill

....... BrdLflcy & Mildicson &

....... 00
.... Mik- 0
kbhd I
NUhd 2
0.1 -
0.1 IM 1000
1 10 (101 (11 10 RX) 1001)

CIVar
1()
: cywlier
li

ä \ý-1 m3

NIKhod 2

0.1 -

0.1 1 10 IOD ]WO

K

Figure 3.17 Methane/air for Predas a function of K

When the same set of data was compared with P, d v AjAv plots, it apparently gave
better trend. The cubic data results seem not given good agreement when A,,/A, term is

applied for V= 30 m.3 All rectangular data fall below Bartknecht's correlation line,

suggesting that the use of A, term is more favourable in the case of non-cubic vessels as
been postulated above. It also showed the same trend when applied to cylindrical vessel.
From the analysis, it can be said that the K term is more favourable when it applied to

cubic and spherical vessels as it fits the spherical flame propagation scaling law. The
satisfactory results on correlating the influence of vessel's shape is given when AJA, is
applied for non-cubic and non-spherical vessels as been proved in Fig. 3.18. It can be
A, instead V 2/3term to get more precise results if the
recommended to use term of
vessel's shape influence have to be taken into account in venting design.
 92

10.0 aj, /air

CylirXier
CK/air A
10.0
A
A'. y
X a. " a.,
X
1.0
X
1.0 XXa
X V-1n13
V-22aG 41
Xa JL
X I a V- 1 rre
44k, x V-30m3 A. BalknaM
X Ba"kiec"
CL 0.1 a ---- Swift
X StAft
a' a B-zideyandlAtMason ....... Bradaya-d Mitchasm
a .... Wkov
AA1 I
a KUlhod2
Wiod 2A
0.0 0.0
0.1 1.0 10.0 100.0 1000.0 al 1.0 10.0 11MO 1000.0
A,lAv A. A

10.00 CFVair
Rellancjd- V- 02 n13
V-0.6 m3
0 V-2.8rn3
1. 0V- 112M3
00 V-20SM3
0* V-26.6m3
/,
Bw4macht
00
Smil
0
1110 Bradley and Mlcrmon
-------
00.... Wdkov
0
kftiod 1
.a08 000
0A M*od 2

(101
0.1 1.0 10.0 100.0 1000.0
NA,

Figure 3.18 Predof methane/air at various geometries as a function of AJA,.

The data shows that Predexponent should be 0.5 in the subsonic flow regime and unity

exponent in the sonic flow regime are supported by the data. Bartknecht's 0.58
exponent is a compromise between the two regimes which not a good correlating
exponent and does not fit the fluid mechanics of the vent flow. The main data scatter is
due to the additional influence of volume. It can be summarised that the proposed

equation (Method 2) for safe venting has more advantages compared to others. The
analysis is based on the mass and momentum balance, only employing an empirical
relationship to account for compressible flow and sonic flow. The use of S,, (E-1) and A,
terms for non-cubic and non-spherical vessels to correlate the important parameters
involved in venting gas explosions seems to agree well with the published experimental
data. The simplicity of this 'safe' venting correlation can be used for any required

venting pressure and area, providing one of the values is given.

 93

3.5.5 Evidencefor an additional influence of vesselvolume on Pred

The datafor eachvolumetestedover a rangeof K was plotted as PredvK and later Prýd
at fixed K of 2,4,10 and 30. This allowedthe influenceof volumeto be investigatedat
constantK. There is insufficient data to undertakethese plots for town's gas and
acetone.Figure3.20 showsthat for methanethereis a very significantvolumeinfluence
for K<4 with no consistenteffect at higher K. The higherK datais mainly in the sonic
flow regime.At low K the increasein Predwith volumeis probablydue to the flame self
accelerationeffect,which extensivelydiscussedin NFPA 68.

It is widely recognisedthat just as in the transition from laminar to a turbulent fluid

flow, flame wrinkling occurswhen the flame Reynoldsnumber(Re) exceedsa certain
value(3000-4000)Le, which is a function of the flame expansionratio, E (Groff, 1982).
In smallervessels,cellular instabilities doesnot appearuntil the late stagesof the flame
developmentjust prior to the vent failure andthus,no significant effect on initial flame
and pressuregrowth at low K. It meansthat in smaller vessel,the critical Reynolds
number is not attained. Further, in smaller vessel, there will be less time for the
couplingbetweenthe acousticwave andthe combustionwaveto acceleratethe burning
rate effectively (Wingerden and Zeeuwen,1983b,Wingerden and Zeeuwen, 1983a).
Unfortunately,for larger vessel, the onset of flame cellularities would occurs in the
early stagesof the explosion,enablingto promoteaccelerationof the flame front and as
a result to attain an increasedrate of pressurerise (McCann, Thomas and Edwards,
1985).It shouldbe notedthat Bartknecht(Bartknecht,1993)workedwith large volumes
(volume ranging from I to 250 rn3 cubic) so the self-accelerationflame had already
occurredto the limit values and this is the reasonwhich the correlation gave non-
conservativepredictionfor A, when applyingto smallervessel.

Further,Groff (Groff, 1982)explainedthatin largervessel,hydrodynamic

instabilities
(transitionof smoothsphericalto polyhedral flame) will arise at somecritical
-cellular
expansionratio. This then acceleratesthe overall burningrate of the gasmixturesowing
to the expandedflame surfacearea,thus increasingthe maximum rate of pressurerise
abovethat to be expectedfrom the laminar system.Hydrodynamicflame acceleration
effects (cell cellularities) produced will increaseKo values in larger vesselsand wall
quenchingin the smallervesselsdepressthe Kc, value.Therefore,the use of Ko values
 94

as a simplescalingparameterfor deflagrationsis not valid (Chippett,1984).Figure 3.19

showsexamplehow Kc;affectson sizeof the vessel.

Roo

too

00

01 "SOW P401

Figure 3.19 Kc3as a function of vesselsize for propane/air(4%) measuredat an initial

pressureof I atm(Chippett,1984).

Larger flamesform cells at a critical diameterand flame accelerationoccurs.The net

effect is an increasein S,,andKr,, which hasnot beenaccountedin Bartknechtor Swift
correlation.In principle the effect is similar to vent inducedturbulenceand could be
accountedby 0 term in the burning velocity equation.Hydrogenbehavesin a similar
way asmethaneat low K andlittle influenceof volumein the sonic flow regimeat high
K asillustratedin Fig. 3.20.For propane,the additionalvolumeeffect is morecomplex.
For K<4 thereis a consistentincreasein Predwith volume,but for higher K there is a
decreasein Predas the volume increasesat constantK. This latter trend is difficult to
explain and in the original Bartknechtdata (Fig. 3.3) showna smaller vent areain the
largervolumes-(30 and 60 M3)for the sameoverpressure. It can be postulatedthat it was
due to the experimentalerrors. The likelihood of explosionsresulting from stratified
fuel-air mixturesis high, especiallywith regardto particularly buoyant or heavierthan

air leaks within structures-,despite this the published literature on this topic is quite
sparse.Current literature on stratified explosionsdeals predominantly with buoyant
gases(DeHaan,Crowhurst, Hoare, Bensilum and Shipp, 2001, Liebman, Corry and
Perlee,1970).
 95

K=2
100 K=4
K=10
K=30
10

l
1 Ai:.....
.........

Methane-air
0.1 llý
0.1 1 10 100
V(ff

K--2
1001
K-4
K=l 0
K--X

e"'
rIr

d
0.1
10 100

V(M)

K=2
10
K=4
X--
, - x- A K=l0
ET x K=30

Hydrogen-air
0.1
10
v(M)

Figure 3.20 Effect Of Predon vessel volume at constant K

In practice the vast majority of accidentalgas or vapourreleaseswill involve a high

molecular-weightcompound which is then likely to lead to the formation of an
inhomogeneous stratified mixture. There has beenlimited work in this area(Tamanini,
2000) andin manyscenariosthis is likely to result in reducedoverpressures.
 96

The effect implies flame decelerationeffect at large volumeswhenthe pressureis high

at large K. This is not an effect that has beenpreviously highlightedin the data. For
sonicventingthe pressuresare higher andcellular flamesoccurearlier at high pressure
andarethereat the endof methane/airexplosionin closedvesselsevenon a small scale,
Andrewsand Bradley (Andrewsand Bradley, 1972)photographedcellular flames in a
300mm diameter closed cylindrical explosion vessel at the end of the flame

propagation.Methaneexplosionsat I bar needa greaterdistancethan 150 mm to form

cells and the pressureeffect causesthem to form at shorterdistances.Thus at high K
and Predwith sonic venting during venting explosion,the self accelerationis likely to
havealreadyoccurredat the smallervolumes.

3.6 Concluding remarks

In this chapter,there is found that the correlation derived by Bartknecht(Bartknecht,

1993)have beenshownto be underpredictedmost of the datapresented.It can be said
that the failure of the Bartknecht'scorrelationis due to the assumptionof the samevent
area is required irrespectiveof the volume, Further, most of the vesselsinvolved in
Bartknecht'scorrelationwere large wherethereis suspectedthat self-accelerationmay

occurduring the ventedexplosionandthis parameterdid not be takeninto accountin an

appropriatemanner.From this work, it is recommendedthat the validity and limitation
of Swift's equation(Swift, 1983)can be extendedto wider rangefor P,,d > 200 mbar
providing the parameterPvis addedinto the equation.

A significant flame self accelerationeffect for subsonicventing was shown for K<-5
and this effect is similar to vent induced turbulenceand could be accountedby the 0
term in the burning velocity equation.The turbulenceenhancementpredictedbasedon
Bartknecht'sequationandproposedMethod 2 were in a good agreementwith 0 derived
from tabulatedexperimentaldatabasedon Method 2 as seenin Table 3.5. It can be
said
that the 0 derivedwas perfectly reasonablevalue for P as usedby other experimenters
(Munday, 1963, Pasman,Groothuizen and Gooijer, 1974, Yao, 1974). The name
experiencesdecelerationeffect in larger volumeswhen pressureis high in larger K and
this effect hasneverbeenhighlightedpreviously. It is postulatedthat at high K and Pred
 97

with sonic vehtingduring the explosions,the self-accelerationis likely to havealready

occurred at the smaller volumes.

It can be suggestedthat the useof K term is more suitableto be applied on cubic and
sphericalvesselfor UD =2 but failed to give satisfactoryresultsfor non-cubicvessels.
The AdA, term is more favourableto correlatethe influenceof vessel'sgeometriesfor
non-cubicvessels.The data and figures shown in this work also illustrated that P,,d
exponentof 0.5 in the subsonicflow regime and unity in sonic flow regime are
supportedby the experimentaldata. The main data scatter is due to the additional
influence of volume. The Use Of Predexponent of 0.582 in Bartknecht's equation
comprornisesbetweensubsonicand sonic flow regime which gave most of the data
scatteredoutsidethe line.

It can be recommendedthat Method 2 gives reasonablygood agreementwith most of

the experimental data and the use of S, term to describe the unburnt gas displaced by
the flame which gives approximately -6 times the mass flow rate suggestsit gives close
estimation on Predin relation with practical application in comparison by with Su term
used in Method 1. The net effect is as Sg is close to the flame speed, Sf in value, the
approach is only slightly lower vent mass flow rate than that based on Sf.
 CHAPTER 4

SIMPLY VENTED GAS EXPLOSIONS: THE

PRACTICAL
 98

4.0 Introduction

In order to further demonstratethe applicability of proposedequations(Method I and

Method2) to thepresentstudy,seriesof experimentalworks werecarriedout using Test
vessel1 and2 asdescribedin detailsin Chapter2.Again, this is the purposeto study the
impact of overpressureon venting at different vessel volume as Bartknecht's work
indicatedthat the samevent areais requiredirrespectiveof the vessel'svolumewhich is
obviouslyincorrect.When the experimentaldataof Bartknechtis examined(Kasmani,
Willacy, Phylaktouand Andrews,2006,NFPA68,2002),it is clear that the exponentof
P,edis fitted to the data with I/K > 0.2 or K<5 and the correlationsshould never be
usedfor higherK. This is not recognizedin NFPA 68 or in the draft Europeanguidance
on venting. The limitation on the correlation is stated as P,,d <2 bar and the
experimentaldatashowsthat this is effectivelya limit of K of 10.However,noneof the
experimentaldata for Pred>0.8 bar fits the correlation and all is lower than the
correlation.This effectively gives the validity of the correlationat K<5 and Pred< 0-8
bar andthis is the incompressiblevent flow regime.Even the work in this areahasbeen
extensivelyinvestigated,yet there is still unclearexplanationon the impact of the vent
burst pressureon the overpressure(Bradley and Mitcheson, 1978a, Bradley and
Mitcheson,1978b,Chow, Cleaver,Fairweatherand Walker, 2000, Cooper,Fairweather
and Tite, 1986,Cubbageand Simmonds,1955,Cubbageand Simmonds,1957,Donat,
1977, Fairweather,Hargrave, Ibrahim and Walker, 1999, Harris and Briscoe, 1967,
Kumar, Dewit and Greig, 1989, Pasman,Groothuizen and Gooijer, 1974, Thorne,
RogowskiandField, 1983).

From previous chapter, it is shown that the published guidelines, particularly with
regardto the questionof appropriatescaling of the correlation parametersare having
their own limitations and validity to use in practise.It is crucial to note that there is
anotherimportant factors i. e. turbulent, self-acceleration,volume and geometryeffect
influencing the mechanism in venting explosion which subsequently giving the
publishedcorrelationa poor agreementwith the experimentalresults. For the effect of
burstingvent on maximum pressurein ventedexplosion,most of the work was limited
to stoichiometricfuel concentrations(Bartknecht,1993,Cooper,Fairweatherand Tite,
1986, Cubbageand Marshall, 1974, Cubbageand Simmonds, 1955, Cubbage and
 99

Simmonds,1957,Donat, 1977,Pasman,Groothuizenand Gooijer, 1974,Runes, 1972,

Zalosh, 1980, Thorne, Rogowski and Field, 1983). Changingthe equivalenceratio
changesthe laminar burning velocity, S,,, of the flame and the massburning rate of
sphericalflamesscalesas SU3. In this work-,the effect of mixture reactivity from (D=0.8
to 1.6 will be investigatedin terms of P.,,,, P, and flame speed,Sf for end and central
ignition. In a ventedexplosionwith a vent burst pressurethe flame is larger when the

vent burstsandthe larger surfaceareagive a greaterburningrate,When the vent bursts,

the outflow of unburntgasis at a higher velocity and so the pressurelossesin the vent
are greater and the influence of the vent burst pressureis significant (Cooper,
FairweatherandTite, 1986).

4.1 Generalfeatures of experimental tests

The geometriesof Test vesselI and 2 usedin the presentstudy were cylindrical vessels
which give the volume of 0.2 and 0.0065m3respectively.The vesselwas closedat the
end andfitted at the other with a circular tube with diameterof 0.162 m, simulatingthe
vent (eitherinitially openor covered)to the 52 m3dumpvessel,which effectively gave
free vent dischargeconditions(its volume was 260 timesgreaterthan the ventedvessel)
but enablethe work to be carried out under laboratoryconditions.The length of Test
vesselI is Im while 0.315 m length is appliedto Test vessel2. For Test vessel1, all
experimentswere ignited either at end wall ignition or at the centre of the vessel
centrelinebut in a caseof Test vessel2, only end wall ignition wasconsidered.Pressure
developmentwas monitored at various locations along the length of the test vessels
(Referto Fig.2.2). The distance,x from the sparkdivided by the test vesseldiameter,D,
x/D of the pressuretransducerswas kept approximatelyconstantfor all test vessels.

For Test vessel1, the,pressurewas measuredat threepositions,Po,P, and P2on the test
vessel and at the dump vessel,pressureis measuredby P6. For Test vessel 2, two
piezoresistivepressuretransducerswere mountedalong the test vessel,namely POand
PI and P6 situatedat the dump vessel.For maximum reducedpressure,P,,,,,,this was
taken from P, pressuretransduceras it locatedat the centreof the vessel for both test
vessels.The flame front motion was determinedusingbarebeadthermocouplesarrayed
alongthe vesseland the tube centreline(symbolsas TI-T3 in Fig.2.2). The flame speed
 100

datawere generatedfrom thermocouples for flame arrival timesoutput and allocatedto

the position midwaybetween the thermocouples or in the caseof the first flame speed
the time between the spark and arrival at the first thermocouple. There was no
in
significantdeadtime the thermocoupleresponse but therewasa large thermal lag due

to the 0.5 min diameterthermocouplebead that was used.This preventedits use for

temperaturemeasurementbut was ideal for time of arrival measurementas the

thermocouplesurvived the explosion. A further advantageof the thermocouple flame

arrival detection technique is that it can be used with hydrogen. The alternative
techniqueof ionisationprobe flame detectiondoesnot work with hydrogen names as
they haveno ionisation.As part of the experimentalprogramme,three,repeattestswere
performed at each condition and these demonstrated good consistency and
reproducibility,with peakpressuresvarying by less than±5% in magnitude.

In Test vessel2, vent coversfrom magazinepaper,aluminiurnfoil, Melinex membrane

and 100g A4 paperwere also usedwith four different bursting pressure(P,) ranging
from 98 to 424 mbar. The vent coverswere located behind the gate valve and will be
burst prior to the ignition for investigatingthe effect of initially closedventing on I'Max.
The reason to use the specific range of P, is due to the P, limitation imposed in
Bartknecht'sequation.Thesedynamicburst pressures,Pdy"wereobservedto be greater
than P, at all tests. The gate valve was closed during the explosion vessel mixture
by
preparation partial pressure. This methodof mixture preparation could not be used
without the exit gatevalve as the vent would burst during the evacuationof the vessel.
The gatevalvewas openingprior to ignition, when the pressurewas I bar on both sides
of the vent, exposingthe vent coverto the explosiongases.The single-holeorifice plate
was usedto simulatethe objective on varying the vent areawhich placed at the end of
the vesseljust beforethe gatevalve asdescribedin Chapter2 (Referto Fig. 2.2).

The presentwork is directedat providing further understandingin thesetwo areasof

gas explosion venting i. e. the impact of overpressureon vessel's volume and the
influence of P, on overpressureon ventedgasexplosion.Bartknecht(Bartknecht,1993)

presenteda vent designcorrelationthat is used in NFPA 68 (NFPA68,2002) and the

European vent design standard (2007). This static vent burst pressure, P, was
determinedusingthe vent in the vesseland slowlY (relativeto the rate of pressurerise in
an explosion)increasingthe pressureof the vesselusing compressedair until the vent
 101

burst.However,the actualburstpressurein eachexplosionwaseasily determinedfrom

the drop in pressureP, whenthe vent burst.

The series of tests involved a range of experimentalconditions in each test vessel.

Different rangeof fuellair concentrationof premixed methane,propane,ethyleneand
hydrogenwerepreparedusing the partial pressuremethodto an accuracyof 0.1 mbar
(0,01 % of composition).The flammablemixture was initiated by an electrical spark
which gives 16 J energiesfor the gas explosiontests.The summary of the tests are
givenbelow,

Table 4.1 Summary of the experimental tests

TEST VE SSEL 1
Fuel/air (D K Vent type Ignition
Methane/air 0.84-1.43 16.4 Open End/Centre
Propane/air 0.8-1.5 16.4 Open End/Centre
Hydrogen/air 0.34-0.54 16.4- Open End/Centre
- Ethy- lene/air 0.6-1.6 16.4 Open End/Centre

TEST VESSEL 2
Fuel/air (D K Venttype Pv Ignition
(mbar)
Methane/air 0.84-1.43 1-0,2.1,3.3, Magazine 98 End
16.4 paper
Aluminium 178
foil
Melinex 209
membrane
I OOgA4 424
paper
Propane/air 0.8-1.5 16.4 Open 0 End
Hydrogen/air 0.34 0.54 16.4 Open 0 End
-
Ethylene/air 0.6 IL 16.4 Open 0 End
-
 102

4.2 Generalexplosiondevelopment

In order to give a generaloverview of the typical explosion developmentof gas/air

pressureand name position recordsfor
mixturesin both vesseltests,two representative
openventingareshownin Fig. 4.1 and4.2 for (D= 1.06for methane/air.

The vertical dashedline is the time at which the flame exited the test vesselmarked as
Tm. From Fig. 4.1, the flame initially propagatedin a slow, laminar phasewith low
pressure and flame speed before the flame approachesthe vent which took
approximately110 ms and 72 ms to travel from the ignition point of end and central
respectively.The averageflame speedis 22.8 m/s for end ignition and 16.5 for central
ignition which was considerablygreaterthan the laminar sphericalflame speedof about
3.0 m/s. Basedupon previousresearchon ventedvessels(Chow, Cleaver,Fairweather
and Walker, 2000, Cooper, Fairweatherand Tite, 1986) and records of additional
thermocouplespositioned in the radial direction of the vessel, the flame initially
developedhernisphericallyfrom the point of ignition at the end wall. Then asthe vented
flow field was set up, the flame beganto elongatetowards the vent and during this
stage,unbumt gaseswere also being ventedout to dump vessel.For centrally ignited,
the flame is initially sphericalflame, progressivelystretchedon one side towards the
vent (flame acceleratetowardsthe vent) andalso at the oppositedirection of vent. From
this explanation,it suggeststhat end ignition had a much larger flame area than the
centralignition and this reflectswith end ignition gavehigher overpressurethat central
ignition. For Fig.4.2, it hasthe sametypical pressure-timeprofile for venting at initially
openvent but the flame accelerates very rapidly towardsthe vent in respectwith the one
illustrated in Fig.4.1. It can be anticipatedthat suction effect is influencing the flame

propagation for this vessel's explosion activity compared to self-acceleration

mechanism.It shouldbe notedthat the vesselvolumeof Test vessel2 is about 30 times
smaller than Test vessel I. This situation has been discussedin detail by McCann
(McCann,Thomasand Edwards,1985).
 10.1)

o.4o 0= 1.06,CH4/ar : T,,, for eficlig0im 0.9

V=0.2 0.8
0.35
End igrition 0.7
0.30 -
CeNreigntion
2' 0.6
cz 0.25
Z9ý T,, for centreigrition: 0E
(D .50
0.20 0.4
U) 0.15 C)
Flarm positim for C:
0.3 V
ernd7igNtion
0.10 0.2
0.05 0.1
ý,Ivj
ý
V.
0.00 0
0(ý 0 0.05 0.1 0.15 0.2 0.25
Tima(s)

Figure 4.1 Typical pressure-time histories for centre and end ignition in 'rest vessel I

O= 1.06,CH44r
0.30 0.14
End igrition TOA
V 0.0065 0.12
0.25

0.10
0.20
0.08 E
0.15
0.06
W Flame position
CD 0.10
0.04

0.05 0.02

O. oo 0.00
0.02 0.04 0.06 0.08 0.1
Tirre(s)

Figure 4.2 Typical pressure-time histories and flame position In Test vessel 2.

As can be seen in Fig. 4.1 and 4.2, the peak I)I-CSSLII'e When tile fl, 1111C
OCCUITC(I ZIII-Cady
exited 1rom the vessel cither when ignited at the end WýIll01-Mltl-ýIl. It is considered that
it is the variation 0[ tile Mass bi.11-11
1-,
ItC ilild I'lanic speed of' the Flame approaching, tile
that has a Strong IIII'luence on the vent I'low and on tile SLIhSCLILICIII
VC111 C011A)LIStioll
hCh,IVIOLn,.
A major feature of' tile explosions Is that there arc suhstantial proportions of'
.
 104

the original flammablemixture in the test vesselafter the flame has exited the vessel
andtrappedat the comerregion insidethe vessel.This is largerfor centralignition than
for endignition asit takeslongertime for the combustionto takeplacebefore reaching
the maximumpressureas suggestedfrom Fig. 4.L The directionof unburnedgas flow,
due to gas expansionbehind the flame front, was preferentiallyin the axial direction
towardsthe vent,wherethe unburnedgasesaredisplaced.The inducedflow through the
vent, aheadof the flame leadsto a significant increasein flame speedsand expansion
ratio in the main vessel.If the ignition is initiated at the end wall of the vessel, it
resultedin an elongationof the flame shapewith a correspondingincreaseof its area
andthus,increasingthe burning rate andflame speedeventually.In the caseof centrally
ignited, the flame will be in spherical vessel initially before progressively being
stretchedon onesidetowardsthe vent andthus,reducingthe flame area.This condition
has beenpostulatedby Ferraraet al (Ferrara,Benedetto,Salzanoand Russo,2006). In
caseof centralignition, there is an indication of higher quantity of residual unburned
mixture in the vesselwhereasalmostcompletecombustionoccurredin the caseof end
ignition that eventuallyleadsto higherpeakpressure.

4.2.1 Resultsand discussionson Test vessel1

4.2.1.1 Maximum pressure,p.,,,, as a function of equivalenceratio

All the graphspresentedin this sectionaretaken as averageexperimentaldatafor open

ventingexplosions.Pm,,,,
is shownas a function of the equivalenceratio in Fig. 4.3 - 4.6
for methane,propane,ethyleneand hydrogenexplosionswith end and central ignition

respectively.

Figure 4.3 showedthat P,,, is higher at all equivalenceratios, (D for end ignition
correspondsto onefor central ignition. The sametrend showedthat end ignition results
higherP.,,, comparedto centralignition for propane/airandethylene/airexplosion to
up
(D = 1.13 and 1.0 respectively.However, the univocal trend is not consistent where
higher P seemsto favour the central ignition for rich concentrationsin
...ý, propane/air
andethylene/airi. e. (D> 1.3.This inconsistenttrendcan be explaineddue to the mixture
reactivity itself Cooper at al (Cooper,Fairweatherand Tite, 1986) reported that rich
propanemixtures have much higher susceptibility than methaneto develop surface
 105

instabilities(flame cellularity) which would lead the flame to self-acceleratecausing

higher burning rates and hencerate of pressurerise and this could result in a more
severeventedexplosionasshownin Fig. 4.4.

0.40 CH4/air
0.35 6 End ignition
Centreign.itio.n.,
0.30
0.25
.01
0.20
cz
E 0.15
CL
0.10
0.05
0.00
0.6 0.8 1.0 1.2 1.4 1.6
Equivalenceratio

Figure4.3 Methane/airat different equivalenceratio, (D

Fundamentally,ethyleneis more reactivity than propaneand thus, it becomesmore

obviouson for
P,,,,,, ethylene/airexplosionfor (D= 1.4 and 1.6 at centrally ignited (refer
to Fig.4.5). The rate of pressurerise shows that at rich mixtures, the central ignition
produceda more severeexplosionand it is significanton propane/air
where dP/dt (max)is
156.6at (1)= 1.375thanin ethylene/airwith correspondingto end ignition on both cases
as clearly pictured in Fig.4.6. The rate of pressurerise is a much more important
yardstick of explosion severity as any explosion protection system (venting or
suppression)hasto copewith the maximumburning rateandmaximumrate of pressure
generation.

It is very interestingto investigatethe more reactive gas mixtures than ethylene i. e.

butaneandpenteneas this gas hasbeengiven little attentionon its behaviourin vented
gasexplosion,
 106

1.8 (ý3Halajr
ig ticn
1.6
Certre I ti
1.4

1.2-,

0.8

0.6

0.4

0.2

0
-1
0.5 0.7 0.9 1.1 1.3 1.5 1.7
Eqjvalence ratio

Figure 4.4 Propane/airat different equivalence ratio, (D

In order to get insight about the reasonof such behaviour,it should be basedon the

residualamountof unburnedgasesleft inside the vesselthat will be discussedin detail

later and the ignition position effect. When mixtures are centrally ignited, the flame is
stretchedon both directions;substantiallypushedout only small amountof burnt gases
from the vessel.It can be postulatedthat combustionis still far form completionasthere
is larger amountof unburnedgasesleft inside the vesselwith respect to the almost
completecombustionif mixturesignited at the end wall ignition which leadsto higher
Pmax-

3.5 C2H4/air

3.0 & Endignition-7

2.5
E"
2.0

1.5
E
CL
1.0

0.5

0.0 ý-
0.5 0.7 0.9 1.1 1.3 1.5 11.7
Equivalence ratio

Figure 4.5 Ethylene/airat different equivalenceratio, (1)

 107

180] C3H./ajr
160 -ý F-- - ---,
ej Endignibon 1
140 Centre grlfioný,
-I-
120-
100-
80-
60
40
20
0---
0.6 0.8 1 1.2 1.4 1.6
Eqd%derr,e ratio

3001 Q2Po/eir

250 A Endl
ceml
200

1501

100

50

0
0.5 0.7 0.9 1.1 1.3 1.5 1.7
EquKderre retio

Figure 4.6 Maximum rate of pressurerise, dP/dt (max) at variousequivalenceratio for

propane/air(above)andethylene/air(below).

6.0 1 H2/air
.6 End Ignition
5.0 41
--. I...

4.0
-cm
Co
m 3.0
;:ý
cu
2,0

1.0

0.0 4--
0.30 0.35 0.40 0.45 0.50 0,55 0.60
Equivalence ratio

Figure4.7 Hydrogen/airat different equivalenceratio, (D

 108

Meanwhile,Fig. 4.7 showsthat centralignition producedmore severeexplosion at all

equivalenceratios for hydrogenlairexplosion.There is no significant in Pmax
on both

ignition positionsfor (D=0.34 and0.41 beforeit gavea rapid increasein Pn,,,,at centre
ignition i.e. Pmax
= 2.7 barg (D
at = 0.54 compared to end ignition at higher equivalence
ratio. Combustion times for central ignition are roughly half of the end ignition,

thereforeless is for
time available gases in the vesselto flow out before combustion is

completed,thereby reducing the effectivenessof venting (Kumar, Dewit and Greig,

1989).This observationimplies that ventingis effectiveat lower H2 concentration( (D<
0.41)but not in higher concentrationin the caseof smallervent areaLe, high K.

It also can be concludedthat end ignition gave higher P,ý,,,,in lean to stoichiornetric
concentration in hydrocarbon/air mixtures and P.. in centrally ignited is highly
dominantcomparedto P.. at endignition in rich propane/airand ethylene/airmixtures
((D= 1.1to 1.6).

4.2.1.2 Flame speeds

Figure4.8 - 4.11 reportsthe averageflame speedsmeasuredin secondhalf of the main

vessel(between T2 and T3 in Fig. 2-2) as a function of equivalent ratio and for the
different ignition positions.It seemsthat end ignition gave higher flame speedsin the
main vesselcomparedto centrallyignited explosionfor methane/airmixtures regardless
of its equivalenceratio. For centralignition, the flame speeds ranged from 2 to 10 M/s
for lean and rich mixtures and up toI5 m/s for the slightly rich mixtures i.e. 0=1.05.
With end ignition, the maximum flame speedis 23 m/s, about 1.5 times higher than
centrallyignition at the sameconcentration.

The faster flame speedwith end ignition can be explained by the fact that the burnt
gasesare only allowed to in
expand one direction which in turn will result in an
elongated flame with increasinglarger surface area (reaction front) and hence faster

expansionthan the centrally ignited flames.This accelerationof the flame towards the
vent is also associatedwith self-accelerationof the hydrocarbon-airmixtures inside the
vesselthroughthe formation of cellularflames.
 109

CH4/ajr
25.017
Endigiltion
Cartrelgiticn
20.0
llao

1
19.0
s
5.0

0.0-
0.8 OL9 1 1.1 1.2 1.3 1.4 1.5
EquivaJenceratio

Figure 4.8 Flame speedsof methane/airat end and central ignition as a function of
equivalenceratio.

The flame self accelerationoccurs after a critical laminar flame propagationdistance

andthe availabledistancewith endignition is twice that with centralignition and hence
self accelerationis more likely. The net effectof the vent dischargeand self acceleration
are for the massburning rate of the flame to increasedue to faster flames, rather than
due to the larger flame areaof the sphericalflame with central ignition. The measured
flame speedis shownas a linear function of the flame distancefrom the sparktowards
the vent in Fig.4.9 for methane,propaneand hydrogen/airmixtures which data were
comparedto NFPA 68 (NFPA68,2002). Figure 4.9 showsthe self accelerationof the
flame, basedon the KG, v. closedvesselvolume in NFPA 68 which showsthat Kr,
-,,,,,
increaseswith vesselvolume.The datain NFPA 68 hasbeenplotted as normalisedKcj,
basedon Kr, for the smallestvolume (5 litres) being that for a laminar flame without
cellularity and a flame speedof 3 m/s appliedto thesenormalisedresults for methane,
3.5 m/s for propaneand 20 m/s for hydrogen/air.This enablesa prediction from the
NFPA 68 data to be made of the self accelerationeffect on flame speedand this is
shownin Fig. 4.9. The first flame speedfor the experimentaldatais basedon the spark
to first thermocoupletime. The developmentof cellular flames in sphericalexplosions
hasbeenobservedto occur at about0.3m andthe flame accelerationdue to developing
cellular flames continues until about 3m (Bradley, Cresswell and Puttock, 2001,
McCann,ThomasandEdwards,1985).
 110

CHAlair
30.0
(U. 1.05
25.0
---- WPA 68 --I

20.0 ia End igrition

--h- Ceftre Igrition,
15.0

10LO

5,0 -jj

0.0
0.0 OL2 0.6 0.8 1.0 1.2 1.4
DsLarm from spark

140.0-1 Vajr
i=0.542
120.0- NFPA 6B
-*-
imo ia End igrition
-,er- Calre Igritionl
80.0-

6(10-

40.0-

20.0-

0.0-90
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Distarce frcm spa* (M)

C, 1
3Pold
,&,; n, WPA 68

40.0- Erd lgrition i

Certre igritionj
35.0-
30.0-
25.0-
20.0-
lao -
10.0-
so-
0.0 W-
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Dstarce from spaiý (m)

Figure4.9 Flamespeedasa function of flame distancefrom the sparkfor openventing.

Bradleyet al (Bradley,Cresswelland Puttock,2001) showedthat the cell sizeschanged

in the flame propagationmechanismat an axial distanceof about 3m for a 550 M3
 ill

rectangularvesselfor hydrocarbon/air.The sphericallaminarflame speedof methane-

air explosionsis about 3n-Ysfor large radius flames,but below the onset of cellular
flame development(Bradley, 1999, Gostintsev,Istratov and Shulenin, 1989, Groff,
1982).

FromFig.4.9, all graphsshowa first flame speedmeasurement at 0.16m from the spark
of around3- 4 m/s and acceleratingrapidly at 0.5 m from the sparkto the flame speed
of 28 m/s for methane/air,42 m/s for propane/airand 124 m/s for hydrogen/air
respectively.This is about5 times higher than the sphericalflame speedbasedon data
excerptedfrom NFPA 68 for thosegases.For centralignition, Fig. 4.9 showsthat the
flame acceleratingis more significant after 0.17 m from the spark.It should be noted
that suctioneffect is 'alsoplayed a significant role in stretchingthe flame towards the
vent,resultinglargerflame areaand massburningrate,thushigh flame speedinside the
vesselin respectto its own laminar burning velocity. The net effect is flame self-
accelerationoccurred at half diameter of the test vessel for hydrocarbon/air and
hydrogen/air.It is also clear that for end ignition the distanceto the vent is greaterand
hencethe flame accelerationcontinuous over twice the distance as for the central
ignition. Bradley et al (Bradley, Cresswell and Puttock, 2001) have shown that the

effect of pressureis to increaseflame cellularity or to reducethe distancefor the first

of cellularflames.
appearance

However,the theory of end ignition gives higher flame speedis not supportedby Fig.
4.10 and 4.13 for propane/airand ethylene/airmixturesrespectively.At (D = 0.8 and
1.0,end ignition gavehigher flame speedsof II and 20 nVsbeforethe flame speedsof
27 m/s attainedfor centralignition comparedto 24 m/s at end ignition at (D= 1.12. As
beenstudiedpreviously,propaneexhibits spontaneous cell structureor flame cellularity
in rich mixtures (Tseng,Ismail and Faeth, 1993,Wingerdenand Zeeuwen, 1983a).In
the caseof centralignition wherethe flame travel is lesscomparedto end ignition, very
little unburnt gas mixture has been vented from the vessel; hence there is still
substantialamountof unburntgasestrappedinside the vesselasillustrated in Fig. 4.11.
 112

30.0 CýFVadr
End igriti
25.0 Certre ig

2D.0

15.0

10.0-

5.0-

0.0 "I- --- -- 1

0.6 0.8 1.0 1.2 1.4 1.6

EqLi\derr-e ralio

Figure 4.10 Flame speedsof propane/airat end and central ignition as a function of
equivalenceratio.

Figure 4.11 demonstrated

plots the time of flame arrival in the comer of the test vessel
at the sparkend,wherea flame arrival thermocouplewaslocated.It shownthat the time
flame leavingthe vesselis aboutthe samewhen flame reachedthe comer region in the
caseof endignition, suggestingthat thereis little amountof flame trappedat the comer
region of the vessel,and hencethe increasein P., will be associatedwith the flame
self-accelerationandflame longertravel distancewhich causesthe elongatedflame area
and thus, higherburning rate and flame speed.However,in the caseof centralignition,
it can be saidthat there is evident showingthat considerableamountof unbumt gases
left inside the vesseland propensityof the cell-cellularity in the case of rich propane
concentrationcausethe rapid rise in the final pressureinside the vessel.The movement
of the flame propagatestowards the vent and at the sametime left some amount of
unburnt gasesat corner region is best illustrated in Fig. 4.12. It can be said that the
occurrenceof externalexplosionresulting from the expelledcombustiblegasesignited
by the emergingflame affects the internal pressureas reportedby Harrison (Harrison
and Eyre, 1987)is not valid in this case.The rapid rise of internalpressureis due to the
rapid combustionof the substantialunbumtgasestrappedinsidethe vessel.
 113

TimeatPrnax -7
1
400 12 RamaleftIhevessel r, Test w-ssel 1
C3FVad
Ramaat the comerregion
WO
x.. Time at Prnax
3M X--- Flameleftthevessel
250 11 Flameat comerregion

2DO -

150-,
4
1001
50
ý- --X-
- .... eK................. .x,
_Erdigrition
Certre Igrition

0.6 0.8 1.2 1.4 1.6

EqLjvalerK>--
rabo

Figure 4.11 Time of flame arrival at maximumpressure,when leaving the vesseland at

the comerregionasa function of propane/airequivalenceratio.

Unburnt grises

lu 0uu0u0
Ignition 00 D0 00 P0 000c0 00
cp 800C
00 CE00 0 ýO "0 00
position
0' 0 0

a000001 8 cl n- 00
co 00 00C. OC8 CP
00 0 00 0808808
C. 0n `4 cp

Figure 4.12 Diagramon the flame movementat end ignition and the unburnt gasesleft
inside the vessel

The pressure-timeprofile of raw test of this concentrationshoweda significant pressure

oscillation towards the end of the combustionprocessas been observedby previous
investigator(Chow, Cleaver,Fairweatherand Walker, 2000) althoughthesedo not give
 114

rise to a distinct pressurepeak. An acousticwave generatedduring this period and

propagatingof flame front by inducedturbulentdue to a suddenventing enhancesthe
unburntgasesleft inside the vessel,resultinghigherburningrate andthus, higher flame
speeds.Markstein has shownthat gas mixtures exhibiting spontaneouscell structures
are very sensitiveto pressurewave and therefore easily exhibit acoustically driven
flame instability (WingerdenandZeeuwen,1983a).

For ethylene/airmixture, centralignition gaveslightly higher flame speedscomparedto

end ignition for all equivalenceratio andreachedmaximumflame speedof 36 mIs at (D
= 1.38 as shown in Fig.4.13. Since ethylene/airis a faster burning velocity mixtures
comparedto propane/air,there would be strong and rapid turbulence induced by the
flame flow towards the vent. Consideringthe valid assumptionof the substantial
amount of unburnt gasesleft inside the vesselin the caseof central ignition, flame
instabilities and acoustic wave pressureinteraction incorporatedwith the effect of

enhancedcombustionwill then result to higher burning rate and hence,the increasein

flame speeds.In this case,flame propagationis suggestedto be influencedby the onset
of turbulent flow, with observedburning velocities of 3.18 to 4.48 m/s being greater
thanthe laminarvalueof 0.68 m/s (Chow,Cleaver,FairweatherandWalker, 2000).

4aO C,2HWajr
35.0 B-d lgrItion
30.0 Certre igpitionj

25.0

20.0

iso

10.0

&0

0.0
-1
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
EqLivcdence
rctio

Figure 4.13 Flame speedsof ethylene/airat end and central ignition

as a function of
equivalenceratio
 115

The flame speedsprofile for hydrogen/air(Fig. 4.14) seemscontradictedwith the

pressureresultsinside the vessel(refer to Fig.4.7). This unstabletrend probablycan be
explainedin termsof combustiontime. For a given set of initial condition and vessel
geometry,combustiontimes for central ignition are roughly half of end ignition. This
meansthat lesstime is availablefor ventingbut not for end ignition. The flame travel
time is longerandthe bum rate enhancement is increasedprior to venting which in turn
increasesthe flame speedsinsidethe vesselaswell asthe self-acceleration.Meanwhile,
for shortertime inducedfor ventingin the caseof centralignition encouragedmore gas
mixture combustionafter the flame left the vesselsincethereis more restrictedflow of
flame to be expelledfrom the vessel.However,the gap betweenthe time when flame
leaving the vesseland reachesthe comer region of the vesselis small as shown in
Fig.4.15, not particularly similar trend in the caseof propane/airin FigAl 1. It can be
postulatedthat acousticoscillation (rapid ventingcan trigger the acousticwave), flame
self accelerationand amountof unbumt gasesleft inside the vesselare the significant
factorsof increasein magnitudeof pressurepeakfor centrally ignition comparedto end
ignition even though the flame speedsshowedopposite results. This brings out the
importantfact that high-bumingvelocity mixture suchas hydrogen/airmixture behaves
differently from thosewith low burning velocities.

90.0 -, FVajr
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0 4-
0.30 0.35 0.40 0.45 0.50 0.55 0.60
EclAvaderce
ratio

Figure 4.14 Flame speedsof hydrogen/airat end and central ignition

as a function of
equivalenceratio
 116

1400 Wad r, Test vesseq 1

1200 0 Tirneed
C3 Ram left the %Arsd
1000 lk Marne Ed caner region'
x.. Time at Pmax
8w -
X. - Rame left the %,
msel
6W 7ý R arre ad comer region
F
400
End igrition
2DD - Centre igrition

0 -7
0.3D 0.35 0.40
...........
0.45 0.50 0.55 0.60
EqLivalance
rafio

Figure4.15 Time of flame arrival at maximumpressure,whenleaving the vesseland at

thecomerregionasa function of equivalenceratio.

4.2.2 Resultsand discussionon Test vessel2

4.2.2.1 Maximum pressure,P...

Test vessel2 is a cylindrical vesselwith internal diameterof 0.162 rn and length of

0.315 in, closedat end and fitted at the other end with an orifice plate at different
diametersimulatingthe vent (either coveredor uncovered)before dischargingto the
dump vesselas shown in Fig.2.3. In the experimentsfor initially covered vent, four
different vent materialswere usedas listed in Table 2.3 andplacedafter the gatevalve.
Only end wall ignition was usedin this seriesof tests.The static bursting pressurewas
98,178,209 and424 mbargfor the vent cover used.

Figure 4.16 showedthe pressuretime history for different gas/air at stoiqhiometric

concentration.It can be clearly seen that it takes shorter time for combustion to
completein hydrogen/airexplosion(t = 0.03s) comparedto methane/airexplosion (t z
0.07 s). It canbe expectedthat whenthe combustiontime becomesshorter,it would be
lesstime availablefor unburnedgasesin the vesselto flow out and hencereducesthe
effectivenessof venting. As might be anticipated,both the timing and magnitudeof all
maximum pressurepeaks is explicable in terms of burning velocity of the mixtures
 117

within the vessel;as the burning velocity increases,the magnitudeof the maximum
pressurepeakincreases,whilst the time of occurrenceafter ignition decreases.It also
appearedthat maximumpeakpressureoccurredafter the flame exited the vesselwhich
suggestedthatthereis the variationof the massbum rate and flame speedof the flame
approachingthe vent which has a strong influence on the vent flow and on the
combustionbehaviour.
subsequent

7.0 TouIH2 I-CH4/air

C3h8/air,
6.0
C2H4/air
5.0 H2/air
:ToutC2H4
.04.0
(D -
TollIC3H8
3.0
(n

2.0
TOUICH4

1.0

0.0

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

Time(s)

Figure4.16 Pressure-time
history for different fuels at (D= 1.0

It is very interestingto note that a suddendetonationspike appearsin the pressuretrace

of H2/airjust after the flame left the vessel,posing a completelydifferent trend with
respectto otherhydrocarbon/airmixturesinvestigated.Figure4.17 showsthat the spiky
pressureoccurswell after the flame hasexitedfrom the test vessel.The deflagration-to-
detonationwasnot expectedto 'appear'in this configuration.As mentionedabove,the
presentstudy concernsvesselwith small length to diameterratios i.e. UD= 2. Taking
into accountthat the explosioninitiated by a weak source(a standardautomotivespark

plug with a 16A there is only a very low probability that a deflagraýtion
to detonation
will occur in a mannersimilar to that observedin elongatedtubes.It can be thoughtthat
the externalexplosionmight causethis phenomenato happen.However,if this occurred
then all the pressuresin vesseltest and the dump vesselwould increaseand eventually,
Fig. 4.17 showsthat this doesnot occur as there is no significant
pressuredifference
betweenpressureinsidethe vesselandpressureinside the dumpvessel.
 118

7.0 1 H2/air ! T..,

6.0

5.0

'a 4.0
m
3.0

(n
2? Z.U
"
a.

1.0

0.0
0.005 0.01 0.015 0.02 0.025 0.03
-1.0 Time(s)

Figure 4.17 Pressuretime history for maximum pressureinside the vessel (PI) and
pressureinsidethe dumpvessel(P6)at stoichiometrichydrogen/air

A major featureof the explosionsis that thereare substantialproportionsof the original

flammablemixture in the test vesselafter the flame left the test vesseland eventually,
theseunburnedgasestrappedin the comer regionswere auto-ignited,as will be shown
later.This is largerfor centralignition thanfor endignition.

Figure 4.18 compares the maximum pressure, P.. of all fuel/air mixtures used in the
tests as a function of equivalence ratio. It showed that the highest Pma,was recorded and

observedat (D= 1.0 for methanelairand propane/airand (1)= 1.18 for ethylene/airand
hydrogen/air.It can be said that the result hasa good agreementto other investigations
reported that at stoichiometric and slightly off-stoichiometric gas/air mixtures,
maximumpressureinside the vesselexhibits higher pressurecomparedto the near and
very rich concentrationas at theseconcentrations,laminar burning velocity is at the
highest values. It is interesting to note the highest P., for hydrogen/air explosion
recordedat 35 % concentrationby volume ((D= 1.18),not at 40 % concentration((D=
1.36) that beenreportedto have highest maximum overpressureand laminar burning
velocity (Andrewsand Bradley, 1972,Andrews and Bradley, 1973,Kumar, Dewit and
Greig, 1989).It shouldbe notedthat all maximumpressureobtainedfrom hydrogen/air
testswererecordedbasedon the peakpressuresnot the suddenpressurespike.
 119

There is large increasein P.. for reactivemixtures i. e. ethylene/airand hydrogen/air

comparedto methane/airand propane/airand reachedmaximum overpressureat 1.25
and2.28 bargfor ethylene/airand hydrogen/airrespectively.Higher Pma,reflectsto the
higherflame speedswhich in turnshighermassburningrate.

3-0 - CH4/edr
--X-
o C3HB/air
25-, C21-Wair
1ý
j --x- H2/är j
20

1.5

1.0

0.5

0.0
0.6 0.8 1.0 1.2 1.4 1.6 1.8 2-0 2-2
Eqdvalercerafio

Figure 4.18 Maximum pressurePmý,.,

for different fueVair mixtures as a function of
equivalenceratio.

4.2.2.2 Flame speeds

The averageflame speedmeasuredbetweentwo thermocouplesfor different gasesis

shown,as a function of equivalenceratio in Fig. 4.19. The flame speedincreasedvery
rapidly in hydrogen/airup to (D = 1.18 before decreasinggradually at the very rich
concentration.It showedthat hydrogen/airexhibited higher flame speedof 53.8 M/s at
(D= 1.18.This is about 28 times higher than 2.1 m/s at (D= 0.34. For ethylene/air,the
peak flame speedof 13.6 m/s occurredat stoichiometricconcentration((D = 1.0); the
sameconcentrationat which methane/airand propane/airexperiencedthe peak flame
speedi. e. 8.2 ni/s and 10.9m/s respectively.The maximumflame speedsof thesegases
 120

wereapproximately2 to 3 times higherthanthe correspondinglaminar flame speedfor

the particularmixtures.Theseflame speedswhich reflect the high overpressureinside
the vesselas shown in Fig. 4.18 can be associatedto thesethree factors; turbulence
insidethe vessel,acousticresonanceinsidethe vesselandcombustionof gasoutsidethe

vessel(Harrisonand Eyre, 1987).As statedearlier,externalexplosionis not the cause

of the high pressureinside the vesselas shownin Fig. 4.17. The possibleexplanation
for this conditionis the inducedturbulentandacousticinstabilities.

The fasterflame speedsreflects higher velocity of unburnedgasesaheadof the flame

that in turns causeshigher turbulencefield at the vent areaas shownin Table 4.2. The
vent can be simulated as a blockage or obstacle which can impede the flame
propagationfrom laminar mode to turbulentflames.When the flame encounteredthis
turbulence,it becameturbulentitself and acceleratedto a maximum speedof 58.3 M/s
in a caseof hydrogen/airat (D = 1.18. In particular, flame propagationthrough the
vesselup until flame front venting is found to be substantiallylaminar with significant
overpressureonly being generatedin the later stages of explosions due to rapid
turbulent combustion in the shear layers and re-circulation zones induced by the
obstacles(Fairweather,Hargrave,Ibrahim and Walker, 1999) but self-accelerationis
not expectedin this test vesseldue to the lengthof the vesselis not on the onsetof self-
acceleration.This can be illustrated with simplecalculationto measurehow big is the
flame area,Af accordinglyto vesselsurfacearea,& For Test vessel2, A,, is 0.2 rn2 and
A, is 0.0021m2.Equationfor Method 2 given in Chapter3 is usedfor this calculation
purpose.Rearrangingthe equationwith densityand compressibilityfactor to be taken
into accountgives;

1/2
1
As
cl,

05
P,,' E 1)
which C, ,.
Cd.e.20"
(4.2)
 121

From the calculation,it is found that the flame areais about40 - 45 % of the vessel
surfaceareafor hydrocarbon/air when the pressuremarked its P,, but only 10 % of A.
a,,
in the case of hydrogen/air. This is the valid explanationfor the occurrenceof
detonationspikeobservedin the testvesselthat will be discussedlater.

Endigribon
60.0-, H,
1v0.0065 rrý 21air

50.0

4aD

30-01

Cl-Wair
10.0

0.0
0.0 0.5 1.0 1.5 2-0 25
EgAvadercerzdio

Figure4.19 Flamespeedsfor the studiedgas/airasa function of equivalenceratio.

The increasein burning velocity ahead of the flame due to turbulence might be
attributedto the greatersurfaceareaof the outer boundaryof the flame. As turbulence
develops,the small scale turbulence begins to influence the mechanismsof flame
propagationand further enhancesthe burning rate (Abdel Gayed and Bradley, 1981).
Basedupon the previous investigationon ventedexplosion (Cooper,Fairweatherand
Tite, 1986),the flame initially developedhernisphericallyfrom the point of ignition at
the end wall. In the presenceof the obstacle(in this case,the vent), the flow sets up a
gradient filed leading to subsequentflame-area increase due to stretching, The
interactionof the flame front and inducedrapid turbulentgenerationassociatedwith the
instability mechanism.However,as beenobservedby McCann et al (McCann,Thomas
andEdwards,1985),cellular instabilities do not appearuntil the late stagesof the name
developmentin smaller vesseland thus, they have no significant effect on the initial
flame and pressuregrowth. The induced turbulent also triggers a rapid combustion of
 122

the substantialamountof unburntgasesleft inside the vesselto undergobulk motion

towardsand away from the vent which resultson increasein burning rate and higher
flame speeds(Cooper,Fairweatherand Tite, 1986, McCann,Thomas and Edwards,
1985).

Table 4.2 Summaryof different gas/airpropertiesat highestflame speeds.Values of

Lewis no, Le and Marksteinno, Ma taken from this sources(Clark and Smoot, 1985,
SearbyandQuinard,1990,Tseng,Ismail andFaeth,1993).

Gas/air Sf(measured),rfl/S S99 ,, IS Le Ma

(D
0.8 Sf
CH4/air 1.0 8.2 6.56 0.91 3.73
CA/air 1.0 10.9 8.72 1.7 3.69

C2H4/air 1.0 13.6 10.9 0.81 3.0

H2/air 1.18 53.8 43.0 - 2.8

43 Denagration to detonation in test vessel

As beenmentionedearlier and demonstratedin Fig, 4.17, the significant pressurerise

eventsor 'spikes'in the pressuretraceswereobservedat a time when the leadingname
front had alreadyleft the vessel.However,this deflagrationto detonationsituation was
only occurredfor hydrogen/airmixture in Test vessel2 as clearly illustrated in Fig.
4.20. This phenomenais also observedin Test vessel1 for both hydrogen/air((D> 0.51)
as shown in Fig.4.21 and ethylene/airat stoichiornetriccondition i. e. (D = 1.0. (Fig.
4.22)

From Fig. 4.20, the onsetof the denotationspike occurredat (D= 0.76 with Sf = 28.6

m/s in hydrogen/airmixtures.Noting that it is a lean concentrationwith Sf < 1970m/s

(Chapman-Jouguet (CJ) flame speed)and the length of the vesselis shorter(L = 0.315
m), the onsetof deflagrationto detonationshouldnot happenedtheoretically (Dorofeev,
Bezmelnitsinand Sidorov, 1995, Moen, Bjerketvedt, Jenssenand Thibault, 1985) but
Fig. 4.20 demonstrated
the oppositeresults.
 123

It clearly shownthat therewasa largedifferenceand indicatesa strongdeflagration-to-

detonationlike event.As shownin Fig. 4.16,4.21 and4.22, the occurrenceof the spike
took few milliseconds-3 ms. Both hydrogenand ethylenecan producefuel/air clouds
which are more sensitive to detonation (Moen, Bjerketvedt, Jenssenand Thibault,
1985).Hydrogen/airmixtureshavethe smallestdetonationcell sizesin comparisonwith
the other commongasesexceptionallyto acetylene.Note that the smallerthe cell size,
the more sensitiveis the mixture.This is basedon the remarkmadeby Ng and Lee (Ng
andLee, 2007).As beenstudiedpreviously(Dorofeev,Bezmelnitsinand Sidorov, 1995,
Moen, Bjerketvedt,JenssenandThibault, 1985,Ng and Lee, 2007), the deflagrationto
detonation mechanismcan be instantaneouslyformed via direct initiation using
powerful ignitersor othermeanssuchas shockfocusing.Alternatively, it can also occur
from turbulent flame acceleration.Rapid turbulent flame accelerationcan lead to
deflagration to detonation transition when sufficiently intense turbulent mixing is
achievedat thereactionzone.The strongmixing and hence,the turbulentcombustionin
the flame zoneis oftenpromotedby the interactionwith obstacles(Ng andLee, 2007).

ao -i H2-air
7.0

ao

510

4.0

ao
20
1.0

0.0 i,
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2-0 Z2 2-4
EqLi\rafercerabo

Figure 4.20 Maximum pressureswith and without spike traces in hydrogen/air

explosionin Testvessel2.
 124

4.0 -1 FVair
1 flý- 1-1+1-

3.51
1
3.01

2.5

2.0

1.51
i
1.0

0.5

0.0
0.04 14
ý
-0.5
Time(s)

Figure4.21 Hydrogen/airexplosionat (D= 0.54 (16 % concentration)for Test vesselI

at endandcentreignition.

0= 1. QC«2Wär
3.5 -

2.5-

,62.0-
:3

CL 1.0-

0.5-

0.0
0 0.02 0.04 0.06 O.OB 0.1 0.12 0.14
TIMM

Figure4.22 Ethylene/airexplosionat (D= 1.0 (6.5 % concentration)for Test vesselI at

end andcentreignition.

Fast flame accelerationtowardsthe vent causesmost of the unbumt gasestrappedat the

comer region of the vesseland both top and bottom of the vessel.As shown in section
 125

4.1.3.2,estimatedhydrogenflame areais 10 % of the vesselsurfaceareawhich means

thereis morebulk flame areainside the vesselduring the combustion,Sincehigh value
of K i.e. small vent area,flow restrictedis experiencedtowardsthe venting of burnt
gaseswhich in turns promote the turbulentjet initiation prior to the vent. A sudden
ventingcan give rise to flame instabilitiesandconsequentlyto more intensemixing of
combustionproductsand reactants(Dorofeev,Bezmelnitsinand Sidorov, 1995). The
oscillatorypeaksobservedfor centraland end ignition in Fig 4.21 could be related to
the inteiaction of the pressure/shockwave with the flame front. The effect of fast
turbulent mixing of hot combustion products with reactant, flame shock interaction and
flame instabilitiescausesthe auto-ignitionof the unbumt pocketsof mixture inside the
vessel or by specific, 'hot spots', leading to the explosion responsible for transition to a
developing detonation. A hot spot or auto-ignition centre comprises a small kernel that
is initially slightly more reactive than the surrounding mixtures (Gu, Emerson and
Bradley, 2003). This argumentis supportedby the time of flame arrival in the comer

region of the Test vessel I at the spark end, where a flame arrival thermocouplewas
locatedas shownin Table 4.3. The resultsshowedthat the time of the flame arrival in
this comerregion was very closeto the time of the pressurespike's occurrencefor end
ignition but eventually not in the caseof centrally ignited. It is consideredthat the
comerregion is not a high turbulencezoneandhence,the auto ignition point or hot spot
is the bestdescribedfor the observation.Similar observationis reportedfor large scale
test using 35 % H2/air concentrationwherethe localisedexplosionoccurredinside the
vesselafter the outflow of flame in venting explosion (Dorofeev, Bezmelnitsin and
Sidorov, 1995). However, this phenomenais not observed to propane/air and
methane/airin both vessels.

Table4.3 Time of flame arrival for Test vesselI

Fuel/air Ignition Time of spike Time at the Time the
position (S) corner region flame left the
(S) vessel(s)
H2/air End 0.08-0.082 0.088 0.077
H2/air Centre 0.09-0.093 0.075 0.051
C2H4/air End 0.067-0.07 0.062 0.054
C2H4/air Centre 0.097-0.099 0.087 0.043
 126

4.4 Influence of vent coefficient, K, volume and burst vent pressure,P, on Pmax

In Test vessel2, vent areahasbeenvariedusingorifice plate with blockager4tio ranged

from 0 to 0.9, which representeda range of K of 1.0 to 16.4. The degreeof vent
coefficient, K can be expressedin term of blockage ratio as,

A= 1
K= (4.3)
A, 1- BR

whereA, is the vent areaand A is the areaof the orifice plate. It is illustrated by Fig.
4.23 that the maximumoverpressurewill be reducedaccordinglywith the size of vent
areai.e. lower K will give lower Pmax in
which this case,Pn,,,,= 0.0178and 0.171 barg
for K =1 and 16.4respectivelywhich about 10timesdifferent in P,,,.. This is due to less
flow restrictionto spell out the unburnedmixtures from the vessel.Since all test has
doneon the initially openventing,there is a largedifferent on P"'a,betweenK=I and
K= 16.4.The maximumpressure,Pmaxof 0.0178achievedin the unrestrictedventing
case(K = 1) happenedat t= 57 ms and interestingly,secondpeak pressureoccurred
later at t= 140 ms. This secondpeak exhibited sustainedoscillatory pressuredue to

acousticpressuredisturbancesgeneratedby fluctuationsin the rate of heatreleaseand

the interactionof the combustionof isolated pocketsof unburnedgas located in the
comer of the vessel(Harris, 1983).The frequencyof the acousticwave is about - 20
Hz. Further,at K=2.1, the sametrend.of this pressure-timeprofile has beenobserved
but not obviousfor K= 16.4.The influenceof the acousticallydriven flame instabilities
appearedto play important role in K=3.3 where two peakspressurebeen observed
which the secondpeak gave frequency of - 143 Hz. For initially open vented gas
explosion,only single peak pressureshould be appearedfor typical idealisedpressure-
time histories(Harris, 1983).This effect hasbeeninvestigatedextensivelyin small and
large vessels(Cooper, Fairweatherand Tite, 1986, McCann, Thomas and Edwards,
1985,Solberg,Pappasand Skramstad,1980,WingerdenandZeeuwen,1983a).

It suggeststhat at high K i.e. small vent area,the egressof combustionproductswere

impeded.The occurrenceof only one peak pressurein K= 16.4
correspondsto the
 127

situation wherethe vent is relatively small and flame speedis higher in which dP/dt
remainspositiveafter the burnt gasesbeenexpelledfrom the vent(Harris, 1983).

0.25 1 CH4/air
o= 1.05

ü2D
K- 16.4

0. e

CLOC

-ao5 lak;*j
1111

Figure4.23 Methane/airat different K for (D= 1.05

As beenmentionedearlier, self-cellularity is not the leading factor of the significant

increaseof P.. from K= I to K= 16.4 obtainedin Test vessel2. It is assumedthat

suctioneffect with the other instabilities mechanismtriggeringthe high P.., in respect

with the increasein K. To justify the hypothesis,predictionon flame areahasbeendone
using Method 2 approachas shown earlier in Section4.1.3.2. It can be said that for
lower K, the predictedflame area is bigger comparedto high K, leading to the high
burningmassandpressurerise, and thusresulting in biggeramountof burnt gasesbeen
expelledfrom the vesseldue to the suction effect. It was shownearlier that the*flame
left the vesselbeforethe peak P.., obtained,suggestingthat combustionis still carried

out inside the vessel.The net effect is lesseramountof unbumt gasesis left inside the
vesselandeventuallywill give lower P.,,,,as shownin Fig. 4.24.
 128

0.18 0-4/air 0.45

0.16 0.40
0.14 0.35
al2l 0.30
CLIO 0.25
0.081 0.20 =
0.06 -i 0.15
0.04
T0.10
aO2 0.05
0.00--, -4(100
0.0 20 4.0 60 8.0 lao lao 14.0 16.0 18.0
K

Figure 4.24 The influence of K on P,,,.,and predictedflame area, Ar for methane/air

mixture.

For hydrogen/airmixture, thereis significantincreasein magnitudeof peakpressurefor

smallervent area(K = 16.4)comparedto largervent area(K = 2.1). From Fig.4.25, it is
clearly shownthat at a very high K, deflagration-to-detonation
transition occurredbut
not at largervent area(small K). However,thereis two peakpressureobservedfrom the
figure, suggestingthat thereis anothermechanismoccurredinside the vessel.At higher
H2 concentration,the initial turbulent inducedby the gas flow, the longer flame travel
time as beenignited at end vessel,the flame front instabilities due to the Taylor and
acousticinstabilitiesaffect the magnitudeof the pressurepeaks.Further,there is a sharp
negative flow from the pressuretraces suggestingthat turbulent from the outflow
flowing back inside the vesselwhich then encouragingthe rate of burning of unbumt
gasestrappedin the comer region of the vesseland hence,increasethe rate of pressure
rise and sharpincreasein secondpeak of maximum pressureas similar as the trend
observedfor methane/airat K=3.3
 129

H2/air
7.0 Tw

6.0

5.0

-ý3 4.0
ACo
3.0

2.0

1.0

0.0
0.01 0.02 0.03 0.04

Time(s)

Figure4.25 Hydrogen/airat stoichiometricconcentrationfor different K.

The V213dependenceof overpressureon the test vesselvolume is a characteristicof

sphericalor compactvessel explosions,where the flame remainsmostly in spherical
shape during venting process. If the spherical flame propagates at a constant rate
irrespective of the vessel volume, there should be no other dependence Of Predon V

other than K. However, there is not what the experiment show in the case of the same K
for different volumesas shownin Table4.4.

From the listed data in Table 4.4, it can be said that in larger volume, 0.2 M3 in the
presentstudy,it is obvious that self-accelerationis the important featurein increasing
the P,,. in Test vessel 1. Comparingthe ratio of Pmjtxj/Pmax2
in Table 4.4 with the
turbulent values calculatedfrom the original Bartknecht's work to the predicted Prd
from Method I and 2 given in Table 3.2, it can be consideredthat those values gave a

good agreement.This implies that the turbulent factor, 0 can be regardedas name self-
accelerationfactor occurred during venting. It can be postulated that the ratio of
PmaxIlPmaX2canbe consideredon how fast the flame acceleratesinside bigger vessel,and
self-accelerationis the main factor of the fast flame propagationandthe high pressure.
 130

Table 4.4 Summaryof experimentalP.. for Test vessel I and 2 for K= 16.4. The
ignition position is end ignition.

Gaslair 4) Test vesselI Test vessel2 Ratio

P.,, j (barg) P. au(barg) ýPmaxi/Pmw
CH4/air 0.80 0.18 0.12 1.50

1.00 0.35 0.19 1.84

1.05 0.34 0.17 2.00

1.26 0.06 0.08 0.75

CjHg/air 0.8 0.14 0.03 4.67

1.0 0.54 0.47 1.15

1.13 0.68 0.30 2.27

1.38 0.35 0.25 1.40

1.5 0.14 0.23 0.61

C2H4/air 0.6 0.04 0.078 0.51

0.7 0.21 0.23 0.91

0.8 0.50 0.72 0.69

1.0 3.06 1.25 2.45

1.4 1.42 1.30 1.09

1.6 0.79 0.40 1.98

H2/air 0.34 0,015 0.027 0.56

0.41 0.11 0.057 1.93

0.48 0.28 0.17 1.65

0.51 0.52 0.25 2.08

0.54 23 0.37 6.21

To further justify that self-accelerationplays important factor in determiningthe final

Pmaxsratio of average flame speed, Sf,,g of Test vessel I and Test vessel 2 was
calculatedandshownin Table 4.5. The flame speedat which the flame front propagates
throughgas/airmixtures during an explosiondeterminesthe rate at which burnt gases
are generated(Harris, 1983).The differencesin the flame speedsshownare due to the
bulk flame areadistortion effects. If Table 4.4 and 4.5 were compared,the ratio value
 131

betweenP,,,,,,,
and Sf,,,s in Test vessel I and 2 were about the samebut significantly
higherfor Sfý,,ratio for hydrogen/air.
g

As shown earlier by using reverse calculation on Method 2, the flame area

correspondingto the Pm.,,is given about 40 to 45 % of total surfaceareaof the Test
vessel2 for hydrocarbon/air.As predicted,for hydrocarbon/airmixtures,it is about 84 -
99 % of flame areaoccupyingthe vessel,meaningthat there is a big cloud of flame
during the explosiondevelopmentif the samecalculationis made.The flame is about
33 to 42 % of total surfaceareain hydrogen/airfor Test vesselI while the flame area
wasjust 10to 27 % of total surfaceareawhen the flameleft the vesselin Test vessel2.
However,in leanconcentrationrange,the flame susceptibilityto cell &ellularity is not
significant(Kumar,SkrabaandGreig, 1987)andthe resultin Table4.5 for hydrogen/air
is not supportedthe argumentmade.It shownthe high ratio of Sfa, Sfavg2 in which can
gl/
be explainedwith themassburningrateof the flameto increase dueto fasterflames,rather
thandueto the largerflameareaand also due to the largerbulk flame left trappedinside
the vesselthat triggering subsequentcombustioninside the vesseland hence,increase
the overpressure
attained.

This finding confirmed the observation reported by McCann et al (McCann, Thomas

and Edwards, 1985) that flame cellularity is appearedin the early stage of the explosion
in larger volume compared to the smaller volume and hence,influence the mass burning

rate and P.,,, inside the vessel. Further, the ratio Of SfavgJ/SfAvg2
is about 2 to 3 for

slightly off-stoichiometric hydrocarbon/air mixtures which suggests that rich mixtures

are known to be more susceptible to develop surface instabilities (flame cellularity)
which would lead to higher burning rates and hencehigher flame speedsattained in Test
vessel I as shown in Table 4.5 in respect to Test vessel 2. This in turn could result in a
more sever vented explosion than might be expected (on the basis of its laminar burning
velocity). This also confirms that vessel volume is the major factor in determining the
final overpressurein vented gas explosion.
 132

Table 4.5 Summary,of experimentalaverageflame speed,Sf,,vl,,for Test vessel I and 2

for K= 16.4.The ignition positionis endignition.
Gaslair 4D Test vesselI Test vessel2 Ratio

Sfavg (M/S) Sfavg (nVs) Sfavgl/Afavg2

CH4/air 0.80 15.51 6.15 2.5

1.00 18-83 8.21 2.3
1.05 22.78 7.51 3.0
1.26 8.35 4.60 1.8
C3H8/air 0.8 11.04 6.15 1.8
1.0 20.01 10.91 1.8
1.13 24.05 8.90 2.7
1.38 15.37 6.32 2.4
1.5 11.89 5.90 2.0
C2H4/air 0.6 6.57 3.41 1.9
0.7 12.25 5.70 2.1
0.8 23.06 11.23 2.1
1.0 28.11 13.61 2.1
1.4 28.61 12.49 2.3
1.6 19.31 7.40 2.6
H2/air 0.34 5.31 2.11 2.5
0.41 22.47 4.78 4.7
0.48 44.69 8.66 5.2
0.51 53.62 10.11 5.3
0.54 85.10 12.68 6.7

As discussedin Chapter3, the pressureat which the vent relief cover, Pvbeginsto fail
has a more significant effect on the magnitude of the P, and P2 (Figure 3.2). The
opening/breakingof the vent would delay or hinder the venting process,causing the
maximum burning rate increasingdue to the bulk flame areacompressedtowards the
vent and hence,the rates of pressurerise increasesas well as the pressureinside the
vesselin comparisonto the openvent mechanism.At low openingvent pressureP,, the
resistanceto flow expansionout of the chamberto the duct is less than at higher P, At
 133

higher P, sincethe vent openedat a relatively late stage,whenthe total flame areahad
increasedsignificantlycomparedto the lower P, case,the rate of burnedgasproduction
exceededthe rate of unburnedgas venting which in turn causeda continuation in
pressurerise within the vessel(Chow,Cleaver,FairweatherandWalker, 2000).

The effect of vent burst pressure,P, hasbeenstudiedextensively(Cooper,Fairweather

and Tite, 1986,Cubbageand Marshall, 1974,Cubbageand Simmonds,1955,Cubbage
and Simmonds,1957,DeGoodand Chartrathi, 1991,Donat, 1977,Rasbash,Drysdale
and Kemp, 1976,Runes, 1972,Zalosh, 1980,Thome, Rogowski and Field, 1983) in
orderto know at what extentthis parameterwill influencethe magnitudeof P.. ', andthe
associatedmechanism.From the investigation,it is found that P.,,,,increaseswith the
increaseof P, dueto the reasonaboveasillustratedin Fig. 4.26.

0.9 -1 e.-Li
-
a8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
0 50 100 ý 150 2DO 250 30D 3w 400 450

P, (barg)

Figure4.26 Influenceof P, on P,,,,,,for Test vessel2

However, this condition is only satisfied the previous works for low K i. e. K =1 and 2.1
but not with larger K in this test configuration. It is apparent that a decrease in P,,,.,,

occurred at K= 16.4 from P, = 209 mbar to P, = 424 mbar, about 1.5 times different
and there was no significant increase in P,,,,,.for P, = 209 and 424 mbar respectively at
K=3.3. It should be noted that the concentration used (10 % methane/air) was the

slightly off stoichiometric which showed the higher P. in the tests conducted at
a.,
varying equivalence ratio. This similar phenomena has been observed in Bartknecht's
 134

work (Bartknecht, 1993) and Harris and Briscoe (Harris and Briscoe, 1967). The
occurrenceof this peculiarbehaviouris dueto two opposingeffects.At higher P, even
if the flow accelerationis stronger,the explosionis vented at a later stage when the
flame is closer to the walls. Thus, there can only be small increase in flame area and

almost immediately followed by gas cooling and flame quenching at the nearby wall.
This results a small increase in burning rate and hence, the overpressure inside the

vessel (Ponizy and Leyer, 1999a, Ponizy and Leyer, 1999b). Similar result has shown in
Leeds work for larger cylinder. Heat losses after the flame has touched the wall make

the peak pressure decreases(Phylaktou and Andrews, 1991, Phylaktou, Andrews and
Herath, 1990). Experiments with turbulent gas mixtures at large vent showed that the

maximum pressure increased rapidly with high pressure of the bursting disc compared
to smaller vent (Harris and Briscoe, 1967). Turbulent could increase the burning

velocity and cause the increase in maximum overpressure. Hence, an explosion vented
in its early stageswill give a higher pressurethan one vented at a later stage.

4.5 Comparison betweentheory and experimental data

Experimental Pmaxat the stoichiometricconcentrationfor Test vessel 1 and 2 was

comparedwith the proposed venting equations(Method 1 and 2) as well as other
correlations, in order for validation as shown in Table 4.6. All calculations used
parameterslisted in Table 3.2 when appropriate. From the table, it shows that
Bartknecht's(Bartknecht, 1993) and Swift's equation(Swift, 1983) offered by NFPA
68 gave grossestimationon maximumpressureand this correspondswith the percent
error deviation betweenthe experimentaldata and calculatedresults as illustrated in
Table 4.7. In comparisonwith the experimentaldata,it is found that prediction given by
Method 2 andBradley and Mitchesonhavea good agreementto eachother. It is due to
the Sg value used on both equations,suggestingthat it best describesthe pressure
developmentin ventedexplosions.For Bradley and Mitcheson'sequation,it
gave 9.5 to
232 % deviationbetweenthe experimentaland calculateddata for Test
vessel I but the
error (168 - 1774 %) is huge for smaller vessel i.e. 0.0065M3 for Test vessel 2.
Meanwhile,Molkov's equationover predictedthe maximum
overpressurecomparedto
experimental data which gave about 189 to 505 % deviation for methane and
propane/airmixture on both test vessels.However, the equationgave better agreement
 135

with experimentalresult on reactivemixturesi.e. ethyleneand hydrogen.This is shown

on lower percenterror deviationof 0.3 to 95 % for both mixtureson test vessels.

As seenon Table4.6 and 4.7, Method I gaveunderpredictionresultscomparedto the

experimentaldata but not for Method 2. Method 2 seemsto have closer results with
experimentaldata whereit gave about 134 to 220 % deviation for Test vessel I and
about127to 489 % deviationfor Test vessel2 (refer to Table4.7). Yet, Method 2 does
not give satisfiedresult when applyingon hydrogen/airmixture wherethe percenterror
is significantly large Le, 1879 %. From this result, it shows that Method 2 gave
consistent agreementwith the experimentalresult for calculating the maximum
overpressure.Among all the publishedcorrelations,Bradley and Mitcheson's equation
fitted well to the experimentaldata for all hydrocarbon/airmixtures but Molkov's
equationgavebetteragreementwith experimentaldataon hydrogen/air.The term S,,(E-
1) usedin Method2 which regardsthe massburningrate to be -7 times higher than the
massburning rate basedon S,,as the flame front rate of consumptionof unburnt gases
seemsto agreewith the experimentalprofile for predicting the overpressurewhich also
been used by Bradley and Mitcheson (Bradley and Mitcheson, 1978a,Bradley and
Mitcheson,1978b) as well as Runes(Runes,1972).The resultsshownin Table 4.6 for
Method I and 2 were calculatedwithout the turbulent enhancementfactor, P. The key
point is that the constantderived from Method 2 do not needany P factor at this stage
(initially open venting)but may haveto be includedfor P, influences.The discrepancy

of addingthe turbulentenhancementfactor, P from previousinvestigators(Bradley and

Mitcheson, 1978b,Chippett, 1984, Pasman,Groothuizenand Gooijer, 1974, Runes,
1972,Swift, 1984)to agreewith the experimentaldatashouldnot be the major problem
in regardsto takeother parametersinto considerationi. e. the densitydependenceon the
pressure,the compressibilityfactor, e andthe orifice dischargecoefficient, Cd.

Again, the sameK resultsthe sameP,,,,,for different volume as shownin Table 4.6 for
experimentalresultsfrom Test vesselI andTest vessel2. It is recommendedto usethe
As term to replaceK in order to provide preciseestimationon venting
area and thus,
Pma,,for given vesselvolumes.
 136

Table4.6 Experimentaldataandcalculatedequationsfor test vesselsat (D= 1.0

Test vesselI
Gas/air Experimental Method Method Bartknecht Swift Bradley Molkov
data 1 2 Eq Eq and Eq
(bar) (bar) (bar) (bar) (bar) Mitcheson (bar)
Eq
(bar)
CH4/air 0.35 0.035 1.12 5.44 12.43 1.163 2.09
C3118/air 0.53 0.04 1.24 7.45 20.92 1.46 2.26
C2H4/air 3.06 0.095 2.84 10.92 20.92 3.35 3.07
H2/air - 1.44 45.13 14.57 - 42.72 4.16
Test vessel2
CH4/air 0.19 0.035 1.12 5.44 12.43 1.163 1.15
C3H8/air 0.47 0.04 1.24 7.45 20.92 1.46 1.36
C2H4/air 1.25 0.095 2.84 10.92 20.92 3.35 2.32
L 1-12/air 1 1 1
2.28 44 45.13 14.57 42.72 4.44
.

Table 4.7 Percent error deviation for experimental and calculation result

Test vesselI
Gas/air Method I Method 2 BartknechtEq Swift Eq Bradley and Molkov
Mitcheson Eq Eq

CH4/air 220.0 1454.3 3451.4 232.3 497.1

-90.0
C31-18/air -92.5 134.0 1305.7 3847.2 175.5 326.4

C21-14/air
-96.9 -7.2 256.9 583.7 9.5 0.3
I I-
H2/air
- - -
I- I
Test vesseli -- -

-81.6 489.5 2763.2 6442.1 512.1 505.3

C3H8/air 163.8 1485.1 4351.1 210.6 189.4

-91.5
C2114/air -92.4 127.2 773.6 1573.6 168.0 85.6

H2/air 1.879.4 539.0 1773.7 94.7

-36.8
 137

Table 4.8 illustratedthe experimentalresults of initially closedventing in comparison

with the calculatedequations.It shouldbe notedthat sincevent cover was in place,the
effect of burstingvent pressure,P, shouldbe takeninto considerationin predictingP,,,,,.,
hencethe turbulenceenhancement factor, 0 is included.For Method I equation,P is 20
and 4 for Method2 equationfor all K exceptfor K= 16.4.From the list, Bradley and
Mitcheson'sequationgave a very satisfiedresult in comparisonwith the experimental
values.Method2 seemsto agreewith experimentaldatafor small K but it gave under
prediction result for K=2.1 and 1.0. Molkov's equationgave bit peculiar results in
for P, at different K. The result show a decreasePliax
order to calculatethe overpressure
at increaseP, which really contradictedwith the experimentaldataobtained.

In summary,it can be concludedthat the experimentaldataobtainedfor simply vented

explosionsarein betweenthe assumptionsof the vent unburntgasmaximum massflow
ratebasedon S,,andS,,(E-1).

Table4.8 Comparisonof methane/airmixture with differentP, for Test vessel2

K= 16.4

P, Exp Pmax PnIax pmx Pnmx Pmax PMOX

(bar) p", (M1) (M2) (Bartknecht (Swift (Bradley (Molkov
(bar) bar bar Eq) Eq) and Eq)bar
bar bar Mitcheson
Eq) bar
I
0.098 0.30 0.59 1.12 5.42 12.43 3.84 0.56
0.178 0.46 0.59 1.12 6.2 12.43 3.84 0.54
0.209 0.82 0.59 1.12 6.5 12.43 3.84 0.53
-, ,
q70- 59 1.12 9.2
I
12.43 3.84 0.48
K=3.3
I-
0.098 0.083 0.38 0.73 0.34 I T 0.5 1.81 0.20
0,178 0.27 0.38 0.73 0.40 0.5 1.81 0.19
0.209 0.47 0.38 0.73 0.43 0.5 1.81 0.19
0.42 0.52 0.38_ 0.73 0.58 1.81 0.17
K=2.1
-098 0.10 0.15 0.29 0.16 0.2 1.33 0.079
0.178 0.36 0.15 0.29 0.18 0.2 1.33 0.076
0.209 0.41 0.15 0.29 0.19 0.2 1.33 0.075
 138

0.424 0.55 0.15 0.29 0.27 0.2 1.33 0.70

K=1 -1

0.098 0.041 0.035 0.067 0.04 0.046 1.07 0.0018

0.178 0.28 0.035 0.067 0.048 0.046 1.07 0.017

0.209 0.37 0.035 0.067 0.05 0.046 1.07 0.17

0.424 0.58 0.035 0.067ý F--607 0.046 1.07 0.16

Table4.8 cont

4.6 Concluding remarks

Experimentsperformedin two different cylindrical vesselvolumesin this presentstudy

have been usedto identify the physics and dynamicsmechanismsresponsiblefor the
generationof the in
pressurepeaks, particular on the of
occurrence the deflagration-to-
detonationin simply vented explosions.In general, end ignition gives higher Pmax
to
compared central ignition in hydrocarbon/airmixtures. The flame is allowed to
in
propagate onedirection, leadingto moreelongatedflame towards the vent andhence,
increasein massburning rate and high flame speed.Self-accelerationis expectedto be
one of the importantfeaturesfor the increasein Pm,,,,
magnitudewhich occurred in Test

vessel I in P,
comparisonwith ax in Test 2
vessel at the sameequivalenceratio. This is
justified by the reversedcalculationdoneusingMethod 2 equation.It shown that about
80-90 % of flame areahasbeenoccupyingthe vesseltotal surfacearea.It is confirmed
the observationreportedby McCann et al (McCann, Thomasand Edwards, 1985). In
their work, they said that the flame cellularity (self-acceleration)appearedin earlier
stage in larger volume and this give significant effect on the overpressureinside the
vessel.For smallervesseli. e. 0,0065 M3, experimentshave shownthat the presenceof
pressureoscillation and this coupling with the induced turbulence by the vent flow
increasedthe P... For centrally ignited explosion, the increaseintensity of the flame
cellularity during flame propagation produces accelerating flame front which later
interact with the vessel wall. Due to the rapid decelerationof the flame front as it
approachthe vesselwall, it results on the strong rare fraction waves which triggering
further combustionof a large amountof unburntgasesleft insidethe vesseland lead to
a significant P.,,, inside the vesselfor reactivegas/airmixturesasshown on ethylene/air
and hydrogen/airmixtures. The ratio Of Pmaxl/PmW
also illustratedthat there was about
 139

2-7 times increasesin overpressurein Test vesselI comparedto Test vessel2 at the
sameequivalenceratio and K, suggestingthat flame is self-acceleratedin bigger vessel
andpropensityof cell cellularity is susceptiblein rich mixturesof hydrocarbon/air.

Auto-ignition is the main factor of the appearanceof spiky pressure traces on

hydrogen/airandethylene/airin the test vessels,It canbe saidthat fast turbulentmixing

of the combustionproductsand reactantsinitiates the 'hot spot' or auto-ignition centre

to leadto the explosionresponsiblefor transitionto a developingdetonation.
 CHAPTER 5

VENTED DUCT GAS EXPLOSIONS

 140

5.0 Introduction

In this chapter,the discussionon ventingexplosionis exploredto the other applications

i.e. venting with the relief duct for gasexplosion.In many practical situations,vented
equipmentis locatedinside a building or nearpersonnelwork areas.For this reason,a
ventshouldbe connectedto the piping ductin orderto dischargethe hot gasesto a 'safe
area'.However,caremust be takenin the designof dischargeducts to ensurethat they
do not compromisethe effectivenessof explosion relief. Studies (Ponizy and Leyer,
1999a,Kordylewski and Wach, 1988, Ponizy and Leyer, 1999b) have shown that
dischargingexplosionproductsinto a duct will alwaysincreasethe pressuredeveloped
in the vesselwhich is being protected-in comparisonto the pressuregeneratedwith no
dischargeduct present.NFPA 68 recommendsthe results of Bartknecht (Bartknecht,
1993)whereP. with a dischargeductsis presentedfor two duct lengthsof lessthan or
ý,,
equalto 3m andgreaterthan 3m. It showedthat about2 to 12 fold increasein reduced
pressure,regardlessof the reactivity of the explosive mixtures in comparisonto the
pressuregeneratedwith free venting,

Works by DeGoodet al (DeGoodand Chartrathi, 1991) and McCann et al (McCann,

Thomas and Edwards, 1985) on varying the duct length in vented gas deflagration
supportedthis observations.It found that the Pna,increasedto more than two fold if the
11
duct length is increased.The surprisingfeaturefrom this resultsshowedthat the length
of the duct is not a major parameterand yet, Bartknechtgaveresultsbasedon his work
for short and long ducts (Bartknecht, 1993). This is becausethe increase in the
is dominatedby duct entry andexit pressurelosses;on which a duct would
overpressure
have to be about 100 pipe diameterslong before the flow friction losseswere equal to
the combinedinlet and exit pressurelosses.The presentwork used a relatively short
duct of Irn long with 162mm-intemal-diameter which gives UD of 6.2 (refer to Fig.
2.3), but it exhibitedall the featuresof explosionventing with a vent duct attached.In
Bartknecht'swork (Bartknecht,1993)the 3m long pipe was 0.5m diameterand hence
the UD was 6, close to that used in the presentwork. As pipe friction lossesare a
function of pipe IJD a dependenceof the vent overpressure the L/D
on would be
expectedratherthan simply on the lengthof the duct, irrespectiveof its diameter.
 141

This is a consequence
of the need to overcomethe inertia of the air within the duct,
beforean outwardflow of gasesfrom the vesselcan be establishedand pressurerelief
commence.In practicetherefore,a dischargeduct should be as short as possibleand
havea cross-sectionalareaat least as large as the areaof the relief vent. It should be
also preferablybe the sameshape,Since evenshort ducts can lead to doubling of the
overpressuregeneratedin the vesselbeing protected,it is suggestedthat whenevera
dischargeduct is usedin conjunctionwith the relief vent, the effective strengthof the
plant should be taken to be half of the value originally assumed.The strengthof the
dischargeduct itself shouldbe sufficient to withstandthe anticipatedexplosionpressure
andthepressureassociatedwith any explosionsthatoccurwithin the duct itself.

Suddenpressurepeakoccurrencein the duct is alsomentionedby Kordylewski and co-

worker (Kordylewski and Wach, 1988) and maximumpressureeffect in the vessel is
found to occur with a particular duct length, equalto about 12 diameters.It is known
now that the intensification of the combustionin the vessel is driven by an impulse
generatedduring the bum-up or secondary'explosion in the initial part of the duct,
shortly after the flame Penetratesinto it (Ponizyand Leyer, 1999a,Ponizy and Leyer,
1999b).Eda et al (Iida, Kawaguchiand Sato, 1985)found that in somecasesthe flame
wasfound to extinguishor hesitatein the channelbeforepassingthrough, dependingon
the equivalenceratio of the mixture, the channelwidth and the flame inflow velocity.
Other studiessupportedthe abovehypothesesby using relatively narrow ducts with a
sharp vessel-ductarea (lida, Kawaguchi and Sato, 1985, Ponizy and Leyer, 1999a,
Ponizy and Leyer, 1999b).When relatively narrowducts are used with a sharpvessel-
duct areachange,the flame front enteringthe duct can be temporarily extinguisheddue
to stretch(causedby the inlet venacontractioneffect, which locally increasescentreline
velocities by 64 %) and cooling through turbulent mixing with unburnedgas which
brings aboutstrongerbum-up (i.e. with higher pressureamplitudes)during re-ignition
(PonizyandVeyssiere,2000).

The presenceof the duct hasbeenaddressedin termsof the increasedpressuredrop due

to the gasflow through the vessel-ductassembly.However,the sameargumenton the
increasedof the pressureinside the vesselis affectedby the violent
explosion occurred
at the initial section of the duct rather than the additional pressurelosses,The flow
restriction in correspondenceof the duct entranceis responsible for a strong flow
 142

accelerationthat produceshigh level of turbulence(Iida, Kawaguchi and Sato, 1985,

Kasmani,Andrews,Phylaktouand Willacy, 2007a,Kasmani,Andrews, Phylaktouand
Willacy, 2007b,Ponizy and Leyer, 1999a).When the flame entersthe duct due to the
high turbulencelevels,hot and freshgasesundergoan effectivemixing that promotesa
violent burning in the initial sectionof the duct. This relatedpressureimpulse in the
duct hasbeensuggestedto temporarilyinduce a flow reversalacrossthe vent (usually

referredto the back-flow).

The only guidelinesfor the designof ductedventsfor gasexplosionsarethoseproposed

by Bartknecht(Bartknecht,1993)which also reportedin NFPA 68. The guidelinegives
barely an empiricalcorrelationbasedon simply ventedvesselsindications presentedin
the samereference.The correlationbasedon the assumptionthat the peak pressurein
the presenceof duct is a function of the peakpressureattainedinside the vesselwithout
the presenceof the duct andthe duct length.The equationsare:

P,,
d with theduct= 0.779 ( Pred
1.013) '-'6'+1.013 Lt<3m (5.1)
withoutthe duct -

Pmd
with theduct -0,172(pred 1.013)*9" + 1.013 3m: 5Uý: 6m (5.2)
withouttheduct -

Equation5.1 and5.2 are the replacementof the previousequationsapplied in previous

version NFPA 68 (NFPA68,2002), the only differencewas the value of the constant
used. The previous equations given for different duct length were provided by
Bartknecht(Banknecht,1993,Bartknecth,1981)asthe correlationsof his experimental
resultscarriedout in aI m3 explosionvesselwith a vent burstpressureof 150mbar and
a vent diameterof 0.2 rn or K= 33.3.The equationsare,

p-red 0.8614
with theduct -I 124 Pred withouttheduct Lt<3m (5.3)

P, 0.5165
d with theduct -2'48predwithout theduct 3m:5 Lt ý!6m (5.4)

Applying all equationsabove,it shownthat the resultswould give overly conservative

predictionby comparingwith the experimentaldataobtainedby Cubbageand Marshall
(Cubbageand Marshall, 1972),McCann et al (McCann,Thomas and Edwards, 1985)
and DeGoodandChartrathi(DeGoodand Chartrathi,1991).For exampleif P, d was 0.4
 143

barg,Eq. 5.1 predictsthat the additionof a vent pipe of the samesize asthe vent would
increasethe pressureto 1.27bar and to 1.04bar for Eq. 5.2. This is about0.56 bar for
Eq, 53 and 1.54given by Eq. 5.4. The presentwork investigatesaK of 16.4 and0.162
m diameter and 1.0 m long vent pipe (LJD of 6.2) for a 0.2 m3 cylindrical vessel.
However,the resultsthat will be shownlater give over conservativeagreementwith Eq.
5.2 and no agreementwith Eq. 5.3 but betteragreementwith Eq. 5.4. Therehasbeenno
validationof Eq. 5.1 and 5.2 for different K or vesselvolumeseven thoughthey are the
basisof the US andEuropeanStandardsfor the useof ductpipes.

The increasein P,,,,,,with the additionof a vent duct is so largethat vent ductscannotbe
used without increasingthe vent area and duct size to achieve a reduction in the
overpressure.However,there is insufficient design data for gasesto enablethis to be
done effectively and the physics of the processfor gas explosion venting is not well
understood.This contrastswith the situation for dust explosions,where a substantial
experimentaldata baseexists (Lunn, Crowhurst and Hey, 1988). Recently, Tamanini
and Fisher (Tamaniniand Fisher, 2003) derived a correlationto take into effect of the
duct when the duct explosionis occurring.All of thesecorrelationsexplicitly take into
accountthe dependence on the duct length,duct diameterandthe vesselvolume.

5.1 Phenomenologyof vesselvented through the duct

Severalstudieshavebeenundertakensincethe beginningof the 1980s,proving that the

presenceof the duct to discharge the explosions products generally increases the
severity of the explosion comparedto the situation encounteredwith a vessel vented
directly to the atmosphere.It has beendemonstratedthat the deflagrationpressurecan
exceedby a factor of 10 or more than obtainedin experimentswith venting directly to
atmosphere(Bartknecth,1981,Molkov and Nekrasov, 1981).This is mainly due to the
complex interactionof the combustioninside the duct pipe and vessel. Despite of the
numerous experimentalstudies and their findings ( Bartknecht, 1993, DeGood and
Chartrathi, 1991, Ferrara, Willacy, Phylaktou, Andrews, Benedetto, Salzano, 2005,
Kasmani,Andrews,Phylaktou,Willacy, 2007, Kordylewski andWach, 1988,McCann,
Thomas, Edwards, 1985, Molkov and Nekrasov, 1981, Ponizy and Leyer, 1999a,
1999b,), the whole picture of the interactionbetweenthe gasdynamics inside the duct
 144

and the vesselas well as the geometry and operating conditions on the maximum
pressureremainsunclear.

Someinvestigatorsproposedthat violent secondaryexplosionoccurredinside the duct

is the important factor leading to an increase in overpressure inside the vessel
(KordylewskiandWach, 1988,Molkov and Nekrasov,1981,Ponizy and Leyer, 1999a,
1999b)while someproposedthat the pressurerise in the vesselis related to the reduced
effectiveness of venting process caused by the pressure rise in the duct
(Ferrara,Benedetto,Salzanoand Russo, 2006). Two different mechanismsfor the

enhanced.burning rate havebeendiscussed.The interactionof the flame front with the

turbulencepromotedby the violent flow reversalfrom the occurrenceof the secondary
explosioninside the duct is one of those severityobservedin venting duct explosions
(Molkov, Baratov and Korolcbenko, 1993, Ponizy and Leyer, 1999a), The other
mechanismproposedis due to the growth of flame instabilities (Taylor-Rayleigh
instability and acoustic instability) triggered by the interaction of the residual
combustion in the vessel with the strong pressurewave produced inside the duct
(KordylewskiandWach, 1988,McCann,ThomasandEdwards,1985).

The additionalfrictional loss due to the presenceof the duct and the duct gas column
inertia are also consideredin reducing the venting flow rate, dependingon the duct
diameter and length (Ferrara, Benedetto, Salzano and Russo, 2006). Bartknecht
(Bartknecth.W, 1981) pointed out that the increaseof the explosion violence and the
peak pressurein the vesselare strongly affectedby the hydromechanicaldrag and gas
column inertia of the duct. The sameobservationhas been describedby Ponizy and
Leyer (Ponizy and Leyer, 1999a)that the frictional lossesand inertia are the key factors
in influencing the evolution of the explosion inside the vessel.This finding has been
confirmedby Ural (Ural, 1993)usingmathematicalmodel.

In this presentwork, it is aimedto investigatethe physic of the venting processwith the

duct attachedto understandwhy the increasein the overpressurewas so large with no
vent bursting pressureand the effect of vent burst pressurewhich in some cases,
reducingthe overpressurecomparedto simply ventedexplosion. There have been few
investigationson the effect of the mixture stoichiometryin venting for the
open vent gas
explosion and none for venting with a vent duct. Changing the equivalenceratio
 145

changesthe larninar burning velocity, S,,, of the flame and the mass burning rate of
sphericalflamesscalesas SU3 This directly influencesthe velocitiesin the ventpipe and
.
henceon the influenceof the vent pipe in the explosion.Thus a study of venting with

reducedburningvelocities will help to understandand model the influence of the vent

pipe in faster explosions(Kasmani,Willacy, Phylaktou and Andrews, 2007). In this
presentwork, the effect of mixture reactivity from (D= 0.54 to 1.6 will be investigated
in termsof P.,,,, flame speedsandunburntgasvelocity aheadof the flame.

5.1.1 Effect of the relief duct fitted to the test vessel

In the previous chapter, the details explanation for the physics and dynamics of the

simply vented gas explosion were given to get better understanding on the venting
mechanism. Figure 5.1 showed the pressure traces for simply vented and duct vented
explosion for methane/air at (D = 1.06 for end ignition. It illustrated that Pmaxfor vented
explosion with the duct fitted is about 4 times higher than for the corresponding Pmaxin
simply vented explosion. A similar trend is also showed for the maximum rate of
pressure rise, dP/dt, It is interesting to note that there are two peaks for dP/dt (ductvenwd)
traces. It shows that the maximum burning rate reaching the first peak after the flame

exited the duct pipe and it suggeststhat the combustion is still taking place after a rapid
decreasein dP/dt to give the second peak before dropping to barg s"' and finally
-80
deceasing to atmospheric condition inside the vessel. The maximum peak pressure for
both casesoccurred at the same time but at different magnitude. This can be proposed

that they have the same mechanism but the presence of the duct has triggering the
pressurerise inside the vessel.
 146

1.6 CH4/ar T, Dict %oled 200

1.4 O= 1.D6
End igrition 150
1.2
1.0
dP/cR SirToyv3"fted 100 m
08 _,
.
0.6 50
0.4
0
0.2
0.0
ý). 04 0.06 0.08 0.1 0.12 0..14 6 0.18 0.2 -50
_().
-0.4 Tin-u(s) -100

Figure 5.1 Pressure-time histories for simply vented and duct vented explosion for
17,
methane/airat 0) = 1.06 for end ignition (initially open venting). dP/dt traces were given
for both cases.

In order to get insight about the mechanism of vented duct eXPIOSiOll, tile flow and

cornhustion dynamic is presented in Fig. 5.2. Again, tile peak overpressure, P,,,,,, at
Pve. OCCUrsafter the flarne has exited the vent pipe, and it happened to all explosions
ýsej
with or without the vent bursting pressure, Pv observed from the experiments. The peak

pressure does not Occur WlIC11the flame is in tile duct as postulated by Lunn et al (Lunn,
Crowliurst and Hey, 1988). After the slow, larninar phase of flame propagation in tile

vessel, the flanic enters the duct (t = ti,,) Which Was followed by a Sudden increase in

pressure at the duct inlet caused by a very fast propagating flarne in the duct (Sjj,,,, = 350

in/s). This fast flame and associated sonic flow condition encounters the strong

tUrbuience field a few diameter of tile vena contracta which leads to a subsequent
violent combustion or known as 111.11-11-up.
The bL11-11-Up
phenomena or known as
secondary explosion lias been investigated
L,
by Bartknecht (BartknectliM, 1981), Poilizy
and Leyer (PonIzy and Leyer, 1999a, 1999h) and Molkov et al (Molkov, Dobashi,
SLIZUkiand Hinino, 1999). They identified the hurn-up phenomena and deduced Owt the

llow reversil resulting froin the action is tile main IIICCIIýInlsfn responsible for tile

increased reiction rate in the vessel.

This strong combustion inside the duct causes the massive reversal of the pressure

(negative pressure drop) towards the test vessel and this marked as AP2 This flow
ý.
pattern is associated with flow pressure (energy) losses which are characteristic of the

geometry and expressed in terms of tile J)"essureloss coefficint, K as heen shown hy

 147

Ferraraet al (Ferrara,Willacy, Phylak-tou,Andrews, Benedettoand Mkpadi, 2005,

Ferrara,Willacy, Phylaktou,Andrews,BenedettoandSalzano,2005)

AP
(5.5)
PS9
2

The pressureloss,AP is measuredbetweenthe main vesseland a point downstreamof

the disturbance.In this case,the pressuredifferenceat the duct entranceis 150 mbar,
marked as AP2-3in Fig. 5.2. The entry pressure loss coefficient is fairly constant
(characteristic for the geometry) and for a large vessel to duct ratio (as in the present

study), it is of the order of 0.5 (Franzini and Finnemore, 1994). The maximum value of
this pressuredifference relating to the induced unburned gas flow was measured just

prior to the flame entry into the duct i.

pipe, e. to the left of the 'Ti"' dashed line in Fig.

5.2. Substituting 150 mbar into Eq. 5.5 using density of p=1.2 kg/m 3 (ignoring any

small pressure rise and any compressibility effects), the calculated unburnt gas flow
is
velocity around 230 m/s. If this calculation is compared to the flame speed inside the
This means that there is a
duct of 350 m/s, the calculated Sgis 280m/s ( Sg= 0.8 Sfiame).

very large flow velocity into the duct pipe and thus will generate a very turbulent flow
field within the duct.

This abruptnegativepressureoccurredin a short duration about t-6 ms but have a

dramaticeffect on the combustiontaking place inside the vessel.It prevents further
outflow of gasesfrom the explosionand at the sametime it increasethe combustionrate
(indicatedby high dP/dt as shownin Fig.5.1) by promotingincreasedturbulence(by the
intensebackflow into the vessel)and by the interactionof the shock/pressurewaves
with the flame front as the flame exits the duct. Consequently,these situation gave a
rapid pressurerise inside the test vesselas shownby P,,,sl tracesin Fig. 5.2 and similar
effect have also been describedby others (Ferrara, Willacy, Phylaktou, Andrews,
Benedettoand Mkpadi, 2005, Ferrara,Willacy, Phylaktou, Andrews, Benedettoand
Salzano,2005,Molkov andNekrasov,1981,PonizyandLeyer, 1999a,1999b)

Positivepressuredropswere thenrestoredand ventingof the explosiongasescontinued.

A pressuredifferenceof >1 bar was establishedbetweenthe vesseland the receiver(i.e.
the dump vessel)and this generatedsonic venting conditionsin the vent. After this the
 148

duct flow canonly increaseits massflow by increasingthe vesselpressure,asthe mass

flow rate in sonic venting is a linear function of the upstreamvessel pressure.The
action of theventburstpressureis to stopany flow in the ductprior to the ventbursting.
The flame propagationinside the vesselwill be longer and the time betweenthe vent
bursting anda flame enteringthe duct will be shorter.The effect of this is investigated
in the presentwork, as it is not clear whetherthis will be beneficial or lead to further
increasesin the overpressure.For very large vent burst pressuresit is anticipatedthat
the higher jet velocities and their suddengenerationwhen the vent cover bursts will
generatemore turbulence in the duct and higher duct velocities will result with
consequentlyhigheroverpressures.

The ignition position clearly has not fundamentalinfluenceon the physicsof the duct
vented explosion,not does the static vent burst pressure,and although both do in
generalincreasethe peak overpressurewith end ignition having higher overpressures
than central and increasing the vent burst pressure usually increases the peak
overpressure.However,in somecasesthis doesnot occur.It is consideredthat it is the
variation of the massbum rate and flame speedof the flame approachingthe vent that
hasa stronginfluenceon the vent flow and on the subsequentcombustionbehaviour.A
major featureof the explosionsis that there are substantialproportionsof the original
flammable mixture in the test vesselafter the flame has exited the vent duct. This is
larger for centralignition than for endignition. It will be shownthat the initial vent flow
reachessonicconditionsandhencethe vent pipe is choked.Principally, the vent flow is
a linear function of the internal vesselpressureand from the relationship,the internal
vessel pressureincreasesuntil the mass of vented gasesreduced the vent flow to
subsonicandlowerpressureloss occur.
 149

T. T.
2.2 A 2.2

1.8 P-. W 1.8

1.4 1.4
A P sý6

1.0 1.0

0.6 Rarre 0.6

porAicn

0.2 0.2

0 02 0 04 O OB O OB ol 0 12 0 1r IO. 2
-0.2 . . . . . .

TmEXs)
AP3ý5
-0.6 AP2ý -0.6

Figure 5.2 Pressure records at selected positions along the test vessel for methane/air at
q) = 1.06 for end ignition. AP2-3IS the pressure difference at the vent entrance, AP3-5is

pressure difference inside the pipe and AP5-6is the duct exit pressure loss. Tin and T,,,,,
indicate the time flame enters and leaves the duct respectively

5.1.2 The influence of static bursting pressure, P, on maximum pressure, P..

As been discussedin detail in Chapter4 previously, the presenceof vent generatesa

pressurewave which interacts with the flame front to distort it and henceincreaseits
surfaceareaand massburning rate. In this manner,the initial flow of gasesinto the duct
will be larger than for an initially open vent. Thesecombinedeffects ofturbulence and
pressurewaves createdby the vent bursting result in accelerationof the flame prior to
the vent duct and also increasethe flow velocity, turbulenceand flame speedin the vent
duct. Testswith ductedexplosion ventsgenerallydisplay Helmholtz oscillations, that is
the pocket of burned gas within the vessel undergoesbulk motion towards and away
from the vent opening, due to the massof the duct contentsand tile compressibility of
the gas in the primary enclosurethat acts as a spring (McCann, Thomas and Edwards,
1985).

In order to investigate the influence of the P, on four different vent covers which
P, ranged between 0 to 500 mbar were used. Comparison results between open venting

and closed venting was also studied to understand the physics of' vented explosion

mechanism on each cases. The variation ol'the maximurn over pressure, Pniax,with static
 150

burstingpressure,P, is shownin Fig. 5.3 with centraland end ignition for (D=1.0 for
propane/airandmethane/airmixtures.

Figure 5.3 shows that all the present results were well below the correlation of
Bartknecht (Bartknecht, 1993) which has been adopted in NFPA 68 (NFPA68,2002)

and the draft European explosion venting guidance. The results at 100 mbar burst
pressure are at least 2 bars below those P,,,. predicted by the design correlation. Figure
5.3 also indicates that the present results for the influence of the vent static burst

pressure showed a lower influence than in the correlation of Bartknecht (Bartknecht,

1993). However, the trends are far from consistent and it is apparent that the effects are
different for propane and methane as well as for end and central ignition.

s CF-WairEdend
14.0-
w CF-Wedr.
1 certre
6 C3HB/airat end
12-0-
?i MHB/dr. Edcerter
0-14, NFPA 68
10.0 C3-IB,WPA 68

8.0

4.0

zo

0.0
0 0.05 0.1 0.15 0.2 a25 0.3 0.35 0.4 0.45
PV(.
M)

Figure5.3 P..,, v P, on stoichiometricpropane/airandmethane/airfor I m.lengthduct.

There arc severalunusualfeaturesin the results.There was a decreasein P.. with Pv

for propaneup to a P, of 180 mbar, Completelydifferent trends for methanethan for
propane were found with central ignition worse for methanethan for propane.The
results with no static vent burst pressureand an initially open vent (Ferrara,Willacy,
Phylaktou, Andrews, Benedetto and Mkpadi, 2005, Ferrara, Willacy, Phylaktou,
Andrews,Benedettoand Salzano,2005) showeddifferent trendsto those with 100mbar
vent burstpressurein relation to the sensitivity of P to the ignition position. This was
 151

much larger with a static vent burst pressurefor propaneand the effect is reversedfor
methaneonce a vent is in position. None of theseeffects are reflected in any vent
guidance(NFPA68,2002)or havebeenreportedby others.

Figure 5.3 also showsthat the influence of P, on Pmaxfor methane/airexplosion was

small for end and centralignition and this behaviourwas similar to initially open vent
ignition (Ferrara,Willacy, Phylaktou,Andrews,Benedettoand Mkpadi, 2005, Ferrara,
Willacy, Phylaktou, Andrews, -Benedettoand Salzano,2005). This trend was also
apparentfor stoichiometricpropane/airwith centralignition. However,propane/airwith
endignition showeda much larger influenceof P, and for all casesthe effect of P, on
P.,,, below 220 mbar was muchmorecomplexthan a linear constantdependenceon P,

For hydrogen/airand ethylene/airtest, the comparisonwith the NFPA 68 (NFPA68,

2002) was not possible to perform at equivalenceratio of 1.0 as they showed a
deflagration-like-detonation
behaviourat (D = 0.57 for hydrogen/airand (D = 0.8 for
ethylene/airand no tests are allowed to perform if Pma,,
exceeded10 bars. Since KC;
values provided by NFPA 68 are given at (D = 1, calculated Ko at different
concentrationsis derived using the relation betweenK,, or Kr' as a function of burning
velocity,S,,(Kumar,Bowles andMintz, 1992),

KG =4.84S(P. IP, )"y(P -Pi) (5.6)

FromFig. 5.4, it is found that Pa, on ethylene/airandhydrogen/airare higher at central

ignition at all P, values comparedto end ignition which totally disagreedwith the
discussionon stoichiometric methane/airand propane/airabove. At end ignition, it
seemsthat the experimentalresults were underestimatedcomparedto NFPA 68 but
inconsistenttrend illustrated for ethylene/airat central ignition. The trend followed the

propane/airresult for P, =0 and 98 mbar but P.,,,,drasticallyincreasedfor P, = 178 and

at the highestfor R, = 209 mbar beforedecreasingat P, = 424 mbar.This observationis
the samefor hydrogen/airwherecentralignition also exhibitedhigher P,,,,.comparedto
end ignition. Thereis no significant different of P. ý,ý observedat Pv = 98,178 and 209
mbar for centrally ignited and only slightly increasedof P.,, at the samePv for end
x
ignition. Interestingly,P.,, is almost the sameat Pv = 424 mbar either for centrally or
endignition.
 152

The possibleexplanationis since the P, is high, so that the time the vent cover is
removed,the pressuredifferential acrossthe vent openingis larger and subsequently,
the pressureinsidethe vesselwill fall rapidly. In fact, the momentumof the gasesbeing
ventedmaybe sufficient to causethe pressureinsidethe vesselto fall temporarilybelow
atmosphericpressure(Harris, 1983).Molkov (Molkov, 1994)also observedthe same
trend in his work. At low P, values(P, < 0.2 MPa), the peakpressureincreaseswith
increasingP, but not at high P, (P, > 0.2 MPa). He explainedthat at high P, the
combustioninside the vesselis almostcompletedand as a consequence,
when the vent
opens,the turbulent 'micro explosion' in the duct does not have any effect on the
turbulenceinsidethe vesseldueto lack of 'Power'.

25.01 --g- Precicted (C2H4)

C2W Et end
C2H4 edcertre
;eu.u ja WPA68 (W)
H2 et end
15.0 H2 at certre
110.0,

.........
..
5.0
0.0
0.00
0.05 0.10 0.15 0.20 0.25
R (barg)
0.30 0.35 0.40 0.45 0.50

Figure5.4 P,,,. v Pvfor ethylene/airat (D= 0.8 andhydrogen/airat (D= 0.54

Whilst at lowerP,, the mixture in the vesselis still not completelyburnedwhenthe vent
opensandthis eventuallyleadinga very strongturbulencein the vesseland combustion
intensity increaseddue to the back flow propagationwhen significant and fast energy

releaseduringthe explosionin the duct. Bartknecht(Bartknecht,1993)also reportedthe

samefinding on his work whereP,,,,,,was increasingwith P, up to P, = 500 mbar and
decreasingP,,,,,observedafterwardsbefore it was increasingagain.This peculiar trend
ý
wasobservedat smallerventareai. e. 0.04 M2for I m3cubic methane/airexplosion.
 153

5.1.3 The influenceof ignition position

In the previouschapter,the influence of the ignition position has been discussedin

details for simply vented explosion which end ignition gave higher P..,, for lean to
stoichiometric hydrocarbon/air mixtures but not in the case of most reactive
gas/mixtures.However,there is no existenceof externalpressurewhich could give a
rise to the internal pressureinside the vessel.In the ventedexplosion with the relief
duct, the interactionof the vesselandthe duct pipe is morecomplexand crucial due to
the confined natureof the external explosion (duct). Ferraraet al (Ferrara,Willacy,
Phylaktou, Andrews, Benedetto and Mkpadi, 2005, Ferrara, Willacy, Phylaktou,
Andrews, Benedettoand Salzano,2005) investigatedthe ventingin the presentvessel
with no vent burst pressure,They found that end and centrally ignited propaneand
methane/airmixturesresultedin approximatelythe samemaximumpressurereachedin
the main vessel,but with substantiallydifferent rates of pressurerise with the end
ignited presenting higher rates. This was attributed to more violent secondary

explosionsin theduct in the end casedueto higher terminal flame speedsin the vessel.
A simple duct flow pressureloss analysiswas used to explain the results, with the
critical event being the entry of the flame into the duct, which causeda dramatic
increasein the overpressuredue to the suddenincreasesin the unburnt gas velocity
aheadof the flame.

From Fig. 5.4 and 5.5, it is obvious that end ignition gavehigher P,,,,, in methane/air
and propane/airat slightly off-stoichiometric mixtures at the presenceof vent cover
compared to the central ignition. It can be elucidated that the different flame
propagationpatternscharacterizingend and centrally ignited n-dxturesare responsible
for different residualamountsof unburnt gasesleft inside the vessel at the time the
flame reachesthe vent and ignites fresh mixtures in the duct. As mentionedin previous
chapter,end ignition gives the maximumdistancebetweenthe flame and the vent. The
action of the vent is to distort the flame shapefrom hemisphericalas the flame develops
preferentially in the direction of the vent, where the unburntgasesare displaced.The
inducedflow throughthe vent duct, aheadof the flame leadsto a significant increasein
flame speedsand expansionratio in the main vessel, known as 'suction effect' in
previouschapter.
 154

In the previous StUdý' (Bradley, Hicks, Haq, I, awcs, Sheppard and Woolley. 2003), the

inf'luence of' flarne stretch will also increase the turhulent velocity with a lower VaILIC01'
L.ewis, Le and Markstein no, Ma inside the vessel. Explosion I'lanies, particularly in rich

hydrocarbon IIlIXtUreS are conducive to tile development of' hydrodynamic and thernio
difl'us,ive effects. At sufficiently small Markstein nuniher, and particularly when they

are negative, the thernio diffusive el'l'cct,, (sinaller Lc,ývis no) are no longer stahilizing.
Thernio-diff-usive instabilities arising 1'rom the prel'erential dil'I'Lision of' reactants with

respect to thernial transport can lead to cellular flanic structures that augment I'lanic

propagation through an increase it) name surt'ace area (Bradley, Sheppard, Woolley,
Greenhalgh and Lockett, 2000). The flame sell' acceleration occurs afier a critical

laminar flame propagation distance and the available distance with end ignition is twice

I that with central ignition and hence self' acceleration is more likely. The net effect of'

the vent discharge and self acceleration are Ior the mass burning rate of' tile flame to

rapidly increase due to faster flarnes in end ignition, rather than due to the larger flatile

area of' the spherical flarne with central ignition. Further, f'roin Fig. 5.4 and 5.5, it is

worth noting that both figures Suggest that the enhanced cornbustion phase in the end
ignition is of very short duration while it stands for a quite longer tirne in centrally
ignited case. The latter case confirmed the hypothesis stated by Ferrara et al (Ferrara,

Benedetto, Salzano and Russo, 2006). They said that in case ot'central ignition, there is

an indication of' higher quantity of residual unburned mixture in the vessel whereas

almost complete combustion occurred in the case c4end ignition that eventually leads to
hig,her peak pressure,

2.5 0=1.06, CH4/ai r End igrition

R, = 178 rrbar
2.0 141
Certre igrýtion
1.5

1.0

0- 0.5

0.0 ý
0.04 0.06 0.08 0.1 0.12 0.14 0.
-0.5 T me(s)

FIgUre5.4 Methatic/air at 1.1,= 178 nibar

 155

4.5 0=1.37S, Cj-Vajr

4.0 P, = 178 rrbar
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
rp. 04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
-().
me(s)
FigUre 5.5 Propane/air at P, = 178 nihar

However, this trend is not rnanifested in a very reactive mixtures i.e. ethylene/air and
hydrogen/air even at the lower equivalence ratio. It shows that central ignition gave
higher peak pressurecompared to end ignition. The same observation was illustrated at
n
high burning velocity mixtures for simply vented explosion previously. It should be

noted that in the present study, the vent coefficient I'm the rig configuration was 16.4
and due to the fast burning velocity ofhydrogen/air and ethylene/air, it creates higherjet
velocities at the vent opening and this sudden onset of the venting leading to the rapid
turbulent at the vena contraction region. In the case of sonic condition (P,,,,,,> 900

inbar), it causes the vent pipe to choke and theoretically, the vent flow is a linear
function of the internal vessel pressure. Since there is higher amount of unburnt gases
left in the vessel for the central ignition in respect of end ignition, it Increases the

cornbustion ratc dUe to the subsequentturhulisation (by the physical back-flow into tile
vessel) and by the interaction ol'the shock pressure waves with the flame 1'rame.Indeed,
the entrainment of' Freshgas pockets operated Lit high velocity hot gases is well known
to originate a violent ignition (Bradley, Emerson and GLI, 2003, Lee and Guirao, 1982)

and under certain conditions, leading to detonation its observed in Fig. 5.6 and 5.7.
 150

[ija r, (1)= 0.54

12.0
P, = 209 rrbar
10.0 r

8.0

6.0

4.0

2.0

0.0
0.04 0.05 0.06 0.07 0.08 0.09 0.1
-2.0
Tme(s)

Figure 5.6 Hydrogen/air at P, = 178 n1bar

velocity mixtures i. e. ethylene and hydrogen,

It can be conclLided that For high bL11-11ing

central ignition showed much largocr influence of' P, on P,,,,,,, contradicting with the

results obtained for hydrocarbon/air. As the reactive mixtures concentration increases,

the combustion time become shorter and less time is available for gases inside the

vessel to flow out before the combustion is completed, thereby reducing the

effectiveness of venting (Kumar, Dewit and Greig, 1989). Tile combustion times lor

central ignition are roughly hall' of' the end ignition and this explained the obtained

results.

8.0 C2Wai r

7.0 Pv ý 424 n-bar

Germ
6.0
5.0
4.0
3.0
2-0
1.0
0.0
0.06 0.08 0.1 0.12 0.14 0.16
_1.00.04
rre (s)

Figure 5.7 EtIlylene/air at P, = 424 mbar

 157

5.1.4 The acceleration of flame towards the vent duct and associated vent

velocities

Kasmaniet al (Kasmani,Willacy, Phylaktou and Andrews, 2007a) showedthat the

inducedflow throughthe vent duct, aheadof the flame, plays an importantrole in the
final severityof the explosion.This flow is driven by the flame speedand burnt gas
expansionin the main vessel. Also, the flame speed was shown to increase with
distancefrom the ignition spark due to self-accelerationthrough the formation of
cellularflame.The self-acceleratingmechanismhasbeendetailedin Chapter4.

In a compactvesselwith LID =2 (the limit of applicationof the Bartknechtventing data

(Bartknecht,1993, Bartknecth.W, 1981)), end ignition gives the maximum distance
betweenthe flame and the vent. The action of the vent is to distort the flame shape
from hemisphericalas the flame developspreferentiallyin the direction of the vent (a

suction effect as shown in Fig. 5.8 by Cooper at al (Cooper, Fairweatherand Tite,

1986)), where the unbumt gases are displaced.The flame progressively stretches
towardsthe vent and this will result in a long cylindrical flame. The datalisted in Table
5.1 (refer to Sf inside the vessel) suggestedthat a much longer flame area (reaction
front) andhencefasterexpansion.At stoichiometricfuel/air mixtures,Lewis no, Le are
1.7 and 0.91 and Markstein no, Ma are 3.69 and 3.73 for propane/airand methane/air

respectively.For ethylene/airat (D = 0.8, the Lewis no and Markstein no are 1.02 and
4.34 respectively(Searbyand Quinard, 1990).However,there is no availabledata for
leanH2concentrationon Lewis no but -0.82 is Marksteinno at (D= 0.54.Markstein no,
Ma is a physicochernicalparameterthat expressesthe responseof flame to stretching
(Bradley,Lau and Lawes, 1992) while Lewis no is the ratio of conductiveto diffusive
fluxes where it describesthe relative rates of heat and mass transfer. Preferential
thermal diffusion is indicated by Le >I and Le <I is a preferential massdiffusion
(Clarke,2002).Lower value of Markstein no with Le <I indicatesthe propensityof the
flame becomeunstableandprone to cellular instability and self-fragmentation(Bradley,
Hicks,Haq, Lawes,Sheppardand Woolley, 2003).The net effect of the self-acceleration
and vent discharge'suction' effect is for the massburning rate of the flame to increase
due to fasterflames.This showsgreaterfor end ignition as the flame travel distanceto
 158

the vent is twice that for centralignition. The increasein flame surfaceareadue to the
'suction' effectand longer flame travel distanceareto compensatefor the lower initial
massburning rate of end ignition with its 50 % reducedflame surfacearea and heat
lossesto the endflange.The net effect from this behaviouris higher overpressurewith
end ignition for methane/airand propane/airmixtures but not in the caseof reactive
mixturesi.e. ethyleneandhydrogenas shownin Table5.1.

In Table 5.1, simple momentum conservation and mass continuity were used to

calculate the downstream duct pipe velocity, Sg prior to the flame entry to the pipe.
Assuming ideal incompressible flow within a duct with no wall friction, mass continuity

gives,

m=pUjA, ý-PUA (5.7)

where m is the massflow rate, A is the duct cross-sectionalarea and U is the flow
velocity. SubscriptsI and 2 refer to the duct measurementat two different point
positions.The gasvelocity is thengiven by,

(5.8)
pA

As the area of the circular pipe is a function of the diameter, we can reduce the
calculationfurtherby,

(A, )
so Sg,,
=
e. el A2
(5.9)

However,Eq 5.9 assumedthat the flame behavesasa piston, which is not the reality as
the flame is 'pear shapedand not flat' (Ellis, 1928).When comparedto the prediction
using the loss pressurecalculationusing Eq. 5.5 previously,the resultedgas velocities
from Eq. 5.9 are very high and greaterthan sonic (in somecasesfor propane/airand

methane/airat high P, but mostcasesfor ethylene/airandhydrogen/airmixtures).

 159

The useof adiabaticexpansion,E assumesthat the reactionwasinstantaneous;with no

heatloss to thevesselwalls andthat the flame front that delineatedthe unburritreactants
from the productsof combustionwasof negligiblethickness.

Mach number,Mach can be usedto stipulatethe flame propagationinside the vessel

The equationgiven as,
to the maximumgeneratedoverpressures,
which correspondent

Mach

wherec,,is thespeedof sound(m/s)which calculatedusing,

ýRT
Co=ý
ME

wherey is the ratio of the specificheatat constantpressureandvolume and takenas 1.4

for air, R is the molar gas constant,T is the temperature(K) and M is the molecular
The Mach no resultedfrom Eq. 5.10 is 0.5 to 2.0 which suggests
massof the substance,
that the sonic flow been experiencedinside the vessel towards the vent and this
condition agreed with the high flame propagation velocities. This high flame
propagationvelocitiescorrespondwith the lower flame residencetime which illustrates
the cooling of burnt gasesproductsimmediatelybehindthe flame front was limited.

Table 5.1 also indicatesthe little changein the peak final flame speedwith P, but it
should be rememberedthat the massburning rate will increasewith P, due to the flame
density term. For the highestvent burst pressure,this gives closeto a 30 % increasein
the massburningrate whenthe vent burstsaccordingto the Kci, t values.This doesnot
e
change the velocity of sound in the vent duct and hence the Mach number is not
affected.A methodof taking into accountthe bulk shapeof the flame is to look at the
rate of changeof P, with time just prior to the flame enteringthe duct. This is shownin
Table 5.1 with Kr,, This shows low values for central ignition at low P,, which
e,,t.
increasewith P, to be nearly 16timesthat for an openvent for methane/airand 10 times
for propane/air,For ethylene/airand hydrogen/air,it is about2 times and 1.4 times that
for an open vent respectively.With end ignition the effect is much lower apart from
propaneat the highest P, This may be due to the greaterinfluence of the increasing
 160

at high P, for spherical celltl'ill 11,1111C

preSSLII-C IVOINI(liltioll,(ILIC10 IIS lill'Lyel'SLII*I',
I(.'C Zil*Cl
ill spite of' tile lower 11,1111C
acceleratiOll.

The faster inass burn ratc approaching the vent as P, ilICI-CiSeSCMISeSSOI1iCFIOWIII the

vent in(] hence choked flow Vý'experiencLA This prevents there heing any oufflow 1'rom
the duct until the pressaire has HSCIIin the VCSSCItO drive the bLIl-Iltgases out. There then
l'ollows a period of' inixed hurnt gas and gas, venlino with micro explosions and
L- 1ý
detonations (propane/air real- ignition at P\ = 427 inhar and most cases in ethylene/air

and hydrogen/air) in the vent duct. This has heen detailed in section 4.2.

Figure 5.9 - 5.10 show the present flame speed meaSUITIllellISwere considerably higher

than the values expected on the hasis of' the development 01' Cellular flames, 1,01,
all
Studied fuel/mixtures in this present work. As been discussed in previous chapter to
tile Occurrenceof self-acceleration of the flame, tile sarne method was applied
COIII*il-l-n
based on the K(;,,,,,,adopted frorn NFPA 68. This enables a prediction from the NITA
68 data to be made of' the self-acceleration effect on flanne speed and this i,, shown in
Fig. 5.9 and 5.10.

25 CH4(NFPA)

C3H8(NFPA)
20
CH4, central, Pv--O
mb
end, Pv= 0 mb
15 -)K-CH4,
C31-18,end,Pv =0
mb
10 C31-18,central, Pv
0 mb
5

0
0.12 0.32 0.52 0.72 0.92 1.12
Distance from the spark (m)

FIgUre 5.9 Flanic spced a function of' tile I'lanic distance from (lie
spark I'm P, l'or
nictliane/air and propane/mr. (1)= 1.0 for hoth gas/mixtures.

Fi,"Ure 5.9 shOWCdthat the present flarne speed measurementsWere

Considerably higher
then the values expecwd on the hasis of the deve1oprnm 01'Cellular flames, for hoth
 161

methaneand propane, However, to

up a flame distancefrom the spark of 0.35m, Fig.
5.9 showsthat the flame speedincreasewith distancewas roughly that expectedfrom
the developmentof cellular flameswith flame speedsof 4 and5 m/s for methane/airand
For
propane/airrespectively. central ignition, the flame speedincreasesto about 12 M/s
for methaneand propaneat close to 0.5m from the spark. For end ignition the large
flame propagationdistancegavea flame speedincreaseto about20 m/s. This is twice
the 10 m/s maximum flame speedrecordedby Bradley et al (Bradley, Cresswell and
Puttock,2001)for self acceleration,
of flamesat a flame 4
radiusof m. The increasein
flame speed was due to the 'suction' action of the vent outflow, as there was no vent

cover bursting effect to create pressurewaves or turbulence.

At the K of 16.4,there was a high velocity createdin the unburnt gasflow through the
vent and this induces movement in the unburnt gas upstream of the vent, leading to
flame accelerationtowardsthe vent.The effect is to distort the flame into a pear shape,
with the apex of the pear passingthrough the vent first. This is
effect similar for

methaneand propane and the difference in the peak flame speedswere not large,

whereasthey should have been large if flame instabilities were the cause of the
acceleration.

It shouldbe notedthat the concentrationof the hydrogen/airand ethylene/airin this test

is 16 % and5.2 % respectivelyi.e. (D= 0.54 and0.8. However,Fig. 5.10 illustrated that
at lower hydrogenconcentration,self-accelerationexisted in the vesselat about 0.5 m.
from the sparkwhich at the half of the vesseldiameter.This confirms the work done by
Kumar et al (Kumar, SkrabaandGreig, 1987)wherethe flame front instability was self-
induced was observed for a mixture 'containing 15 % hydrogen in air. It is also
interestingto note that the considerableflame accelerationhad taken place before the
vent for end ignition to increasethe flame speedto 43 m/s. It also seenthat the flame

speed is laminar for about 6 m/s at 0.36 m from the spark at end ignition for

ethylene/airandhydrogen/airandthis trend of the laminar flame speedat this point also

beenobservedin methane/airand propane/airmixtures in Fig.5.9, Again, there is about
1.5 to 2 times increase in flame speed of hydrogen/air compared to ethylene/air,
propane/airandmethane/airfor both ignitions. For central ignition, the maximum flame
speedis about25 m/s for bydrogen/airand 17 m1sfor ethylenelair.
 162

Fig. 2a. Fig. 2b.

fig. 2c. Fig. 211.

fig. Ze. Fig, 2f.

Figure 5.8 Reproducedfrom Cooperet al (Cooper,Fairweatherand Tite, 1986).Flame

propagationmechanismat the variousphasesof a ventedexplosion.
 163

Table5.1 Explosiondatafor 4.0% propane/air,9.5%methane/air,5.2% ethylene/airand

16%hydrogen/airmixtures.Flame speed,Sf in the vesselrefersto the speedfrom the
sparktowardsthe vent (TV4-TV2)while flame in the duct is the averagevalue obtained
along the duct (TO,,,
-T,,).S, is the turbulent burning velocity as Sf/E where E is the
adiabaticexpansionratio. Kr, is (dP/dt) V1/3and KGM'ax
is basedon the maximumdP/dt
from pressuretransducerPI, which occursafter the flame hasexited the duct. KGvent
is
basedon the rate of pressurerise just prior to the flame enteringthe duct pipe and is
proportionalto the massburning rateupstreamof the vent.* Spiking pressuretrace
Datafor 4.0%propane/airand9.5%nietliane/airmixtures
Sa.
duct

=Sg Unburnt gas

so (0.510 velocity prior
St =Sr 162)2 to duct
.
(E- entrance
Vewl.
Sr 1)/E 9.52 Sr Sgduct
=Sý
PV PMOX Ko E Sg duct, nVs
Vessel
Fuel Ignition (barg) (barg) Kr,,,.
Verg M/S M/s nVs M/S rn/s
C31-18 Central 0 2.1 67.3 0.5 10 2.0 8.0 76 171 137
C31-18 End 0 2.3 137.4 5.9 37 7.4 29.6 282 227 182
CH4 Central 0 1.3 63.7 0.3 11 2.2 8.8 84 112
140
CH4 End 0 1.4 97.1 4.3 19 3.8 15.2 145 159 127

C3Hg Central 0.098 0.7 57.9 0.5 20 3.6 14.2 135 61 49

CA
End 0.098 1.6 105.3 5.1 27 4.9 19.6 187 134 108
CH4 Central 0.098 1.9 88.3 1.5 16 7.6 30.4 289 156 124
CH4
End 0.098 1.3 95.9 3.5 24 3.8 15.2 145 105 84
C3H8
Central 0.178 L2 91.8 , 4.8 24 4.0 16.0 152 225 180
C31is End 0.178 2,5 150.3 5.5 29 5.7 22.9 218 323 259
CH4 Central 0.178 2.2 94.7 5.0 25 8.0 31.8 303 250 200
CH4 End 0.178 2.1 139.0 3.8 27 5.0 19.9 189 179 143

C3118 Central 0.209 1.8 89.5 5.3 20 5.5 21.9 208 237 189
C3Hs End 0.209 2.8 201.2 5.5 35 12.4 49.6 472 431 345
CH4 Central 0.209 2.5 141.5 5.5 22 6.7 26.8 255 257 205
CH4 End 0.209 2.2 142.7 4.5 24 4.7 18.6 177 135 108

C3148 Central 0,424 2.9 117.0 5.5 30 5.2 20.8 198 216 173

4.0

*6.5

C3Hs End 0,424 296.5 5.5 55 20.2 81.0 771 666 533
CH4 Central 0.424 2,7 109.4 5.5 26 7.9 31.7 302 313 250
CH4 End 0.424 2.5 19011 5.5 34 5.5 21.8 208 290 232
 164

Data for 5.2% ethylene/air and 16% hydrogen/air mixtures

Fuel Ignition P, Plux KG KG S, S, = Sg Sg Sr Unburnt
(barg) (barg) (vent) (,., ) SI/E =Sf (duet) fduct), gas
nVs M/S (E- ni/s nVs velocity
prior to
Sg duct
(0.5 entrance
M/5 /0.1 S9
(duct).

62)2 Tfl/S

9.52
Sg
C2l14
Central 0 *7.7 469 2.7 17.1 3.4 13.6 130 181 187
CA
End 0 *5.0 323 7.9 16.8 3.4 13.5 128 272 331

H2 Central 0 *3.4 190 11.6 25,2 5.0 20.2 192 227 330

H2 End 0 *3.5 142 7.9 21.4 4.3 17.1 163 227 208

C2144 Central 0.098 *7.3 502 0.3 11.7 2.3 9.4 89 211 272
CA End 0.098 *6.1 364 8.2 17.2 3.4 13.7 131 156 296
H2 Central 0.098 *9.9 323 6.4 25.9 5.2 20.7 197 453 306
H2 End 0.098 *4.2 232 4.4 28.5 5.7 22.8 217 243 404

C2H4 Central 0.178 *9.7 580 1.8 15.2 3.0 12.1 116 234 385

CA End 0.178 *6.2 416 7.3 19.8 4.0 15.8 150 190 404

H2 Central 0.178 *8.8 354 12.4 27.4 5.5 21.9 209 509 410
H2 End 0.178 2.8 133 2.0 28.3 5.7 22.7 216 298 435

C2114 Central 0.209 *10.7 650 3.8 15.7 3.1 12.5 119 247 408
C2114 End 0.209 *6.9 480 7.4 22.8 4.6 18.3 174 235 459
H2 Central 0.209 *9.1 312 13.6 31.5 6.3 25.2 240 544 420
H2 End 0,209 3.2 181 10.5 30.5 -6.1 24.4 232 338 439

C2H4 Central 0.424 *7.4 461 5.1 10.0 2.0 8.0 76 252 467
CA 24,2 499
End 0.424 *6.9 519 7,7 4.8 19.4 184 243
H2 Central 0,424 *8.9 331 16.5 17.3 3.5 13.8 132 319 453
H2 End 0,424 2.0 104 5.7 37.2 7.4 29,8 284 341 499

Table 5.1 cont

 165

50 C
--0- 2H4, end, Pv -0mb
45 --[4-C2H4, centre, Pv-Omb

40
35
30
25

20

a: 15

10
5
U
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Distancefrom the sr)ark Im)
Figure S.10 Flame speed as a function of distance from the spark for P, =0 for

ethylene/air ((D = 0.8) and hydrogen/air ((D = 0.54). Noted that only hydrogen line for
NFPA 68 was drawn as there is no available data for ethylene in the guidance.

To justify the influence of vent cover beenplaced during the presentwork, Fig. 5.11
and S.12 shownthe upstreamflame speedresultsfrom the experimentaltests.The mass
of the burnt massat the time of the ventburstingmay be calculatedfrom the assumption
that the pressurerise at the time of the vent burstingoccurswhenthe burnt massof gas
is P,/Piniti,,
1, wherePiiti,,,is the initial of
pressure I bar atmospheric If
pressure. the flame

shapeis assumedto be a spherefor centralignition and a hemispherefor end ignition

and the volumeof burnt gasis assumedto be the massburnt times the unburntto burnt
gas density ratio for constant pressurecombustion (7.6 for propane and 7.5 for
methane),thenthe flame radiusmay be calculatedwith the further assumptionthat the
flange and sparkplug heat lossesmay be ignored. Thesepredictedflame radius at the
point of the ventburstingareshownin Table 5.2.
 166

Table 5.2 Predictedflame radiuswhenthe vent burstfor gas/air.

Vent burst 98 178 209 424

pressure
(mbar)
Methane/air 0.16 m 0.20m 0.21 m 0.27 m
(Central
ignition)
Methane/air 0.21 m 0.25 m 0.27 m 0.34 m
(End ignition)
Propane/air 0.16 m 0.20 m 0.21 m 0.26 m
(Central
ignition)
Propane/air 0.20 m 0.24 m 0.26 m 0.33 m
(End ignition)
Hydrogen/air 0.15m 0.18 m 0.19 m 0.24 m
(Central
ignition)
Hydrogen/air 0.19 m 0.23 m 0.24 m 0.30 m
(End ignition)
Ethylene/air 0.14 m 0.17 m 0.18 m 0.23 m
(Central
ignition)
Ethylene/air 0.18 m 0.22 m 0.23 m 0.29
-m
(End ignition)

The reliability of thesepredictionsmay be judged by comparisonwith the position of

the flame determinedfrom the flame position as a function of time in Fig. 5.11- 5.14.
Hydrogen/airat end ignition, predicted0.3 m flame radius gave good agreementto the
measuredflame position when the vent burst was 0.34 m. For methane/airwith central
ignition at P, = 424 mbar,the measuredflame position when the vent burstswas 0.3 m,
very close to that predicted( refer to Fig. 5.14). For end ignition with propane/air,the
measuredflame positionsin Fig. 5.12 and S.13 were0.3 and0.4m for 178and424 mbar
respectivelyandthesewereroughly 20% greaterthan that predicted.
 107

12 Tan 1.8

10 1.5

8 1.2

6 0.9

4 0.6 E
cl-
2 0.3

0 0

-2 -0.3

Fi,-Ure 5.11 Hydrogen/airl at 1) = 0.54 foi- end 11,= 209 inhar

The reason for the greater flame movement than expected with end ignition was (ILICto

the greater expansion of the flame on the centrelinc compared with the edges, which
were subject to more cooling. This is a well known effect Coi large L/D vessel and for
very large L/D vessels results in an elongated U flame, with very fast centreline flarne
speeds(Phylaktou and Andrews, 1991). If' the distances in Table 5.2 are compared with
the flame speed results in Fig. 5.15 it can be seen that is soon as the vent burst there

was in increase in the flame speed. This occurs at a shomer flame radius for central
ignition, as predicted in Table 5.2. Also Fig. 5.9 and 5.10 shows that without a vent

cover, flarne acceleration above that (We to cellular flames, started at distances similar
to those in Table 5.2 for P, = 98mbar.
 I(

3.0 tn tout 1.8

25 1.5

2-0 1.2

0.9

1.0 0.6

0.5 0.3

0.0 0.0

-0.5 -0.3
0 0.05 0.1 0.15
Time(s)
Figure 5.12 Propane/airat (1)= 1.0 for end ignition. P, = 179 n1bar

7.5 tn 1.9
1.7
6.5
1.5-
5.5
1.31
4.5

3.5 0.9

2-5 0.7
0.5
1.5
0.3
0.5 0.1

-0.5 -0.1
0 0.04 0.08 0.12 0.16
Time(s)

cild ignition. P, = 424 nihar. Spiky pressure peak-

Figure 5.13 Propane/air at (1)= 1.0 1'()1-

is ohserved in this test.

 109

3.5 2.1
tin totil P2
3.0 1.8
P,
2.5 1.5 E
P.)
-3
2.0 Pduct 1.2
1.5 P 5-6
.2
U.9- (-J)-,,,
0 CZ
1 Flame position P3-5 0.6 cc/)L
-0
0.5 0.3
U-
0.0 X- 0.0
1
-0.5 I! 1 -0.3
0.05 0.1 0.15
Time(s)

FigUre 5.14 Methane/air for (1)= 1.0 at central ignition. P, = 424 illbar

As mentioned earlier, the I'lame speed after the, vent burst increases rapidly for both

central and end ignition as shown in Fig. 5.15 and 5.16.

30 C1War, Erd igrition
CFVar, Certre igribon 40
Pv -0 ý- N- Onb
ý- rrb
PV- G8 M
35
25 -44- PV- 9Být' F-ý 176rrt)
Pv- 178ýb
Pv - 209 ýt, 30 K- 2CE4 ýb
20
Pv . 4P4 ýb --W K- 42.4rrt)
25 CkWýýA 6H)
clii(WJ -A GBI
(WTIA68)
15 r-I ji (r+ PA

15
10
1,
5
---W- 5- -LL

00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0.0 0.2 0.4 0.6 0.6 1.0 1.2 1.4
Disfatý trcmftý spark (m) Di starce from ft sparx (m)

35,0 G3FVar, GeWeigrition

--e- Hv-0 IM)
60 Qjl-6/ar, Erd igriban
-G-N-aHnY.
Pv ý 98 ýt
I- 44- 30.0 PV- 19H'A'
Pv - 17B"t] I
50 x Pv-209mb
N-20k) It, 424 mt,
25.0 -* -N-
f FIV- 424
40
ý ("OPA68 ) --* - D44 (f,4FPAfi8)
20 0 ; -*- G3HR (NFPAGO)
-, Z" (" PA (41) .
30
5.0

IT (r 0.0

x-5.0

10
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
0.0
0.0
x

0.2 0.4 (16 0.8 1.0 1.2 1.4

[Astamýt lrc-Anthaspaik(rn)

Figuire 5.15 Influence of Pv on flanlC SIVed LIPSti-CIlll01'thC VCIIJf0j' jjjetjjjjje/, jlj' ((I)
1.0) and propane/ali-((I)z--1.0) with end and central ignition.
 170

For the samedistancefrom the spark,the sphericalflameswhich are assumedto behave

in the centralignition are faster and the fast accelerationstarts at close to 0.2 in, as
predictedin Table5.2. For endignition the fast accelerationstartslater at 0.36 m for all
gasesand all burst static pressuresas illustrated in Fig. 5.15 and 5.16. There was
insufficient spatialresolutionof flame speedto determinethe predictedlarge radius of
the flame whenthe ventbursts,that arepredictedin Table 5.2. The reasonfor the faster
accelerationof the centralsphericalflame whenthe ventburst wassimply that the flame
is closer to the vent and influencedby the vent flow soonerthanfor end ignition. When
the central sphericalflame is 0.2m.from the spark,it is only 0.3m from the vent. With
end ignition, the flame would haveto be 0.7m from the sparkto be 0.3m.from the vent.
Both Fig. 5.15 and 5.16 showedthat at this distancefrom the vent P, startsto have a
significantinfluenceon the flame speedandthis increaseswith P, This is a very similar
effect for centraland end ignition. The influence of P, amplifiesthe flame speedsmore
than ten timescomparedto laminar flame speeds.Further,the increaseof flame speeds
could also dueto the accelerationmechanisms:generationof flow turbulenceupstream
of the vent burstsand the associatedpressurewave which can interact to distort the
flame front andhenceaccelerateit.

However, for end ignition there is a period of flame accelerationbetweenthe flame

position whenthe vent bursts,which Table 5.2 predictsto be between0.2 and 0.34 m
dependingon P, and the 0.7m position when the flame is under the influence of the
vent outflow velocity or 'suction' effect. During this flame propagation period of
approximately 0.36 - 0.5m (decreasingas P, increases)the flame continues to
accelerateby expansionon the centreline,developinga U-shapedflame. Phylaktouand
Andrews (Phylaktou and Andrews, 1991) measured the flame velocity to be
approximately10 m/s at an X/D of 1.5 in a tube of larger L/D and Figs. 5.15 and 5.16
show that this is similar to the flame speedmeasuredin the presentwork at the 0.7m
position (X/D = 1.6). It is. an extra length for fast U-flame development prior to
encounteringthe accelerationeffect of the vent outflow velocity that makesend ignition
have a higher flame speedat the point of flame entry into the vent duct. This then
producesthe higher overpressuresshown in Table 5.1. There is no need to invoke
turbulenceeffectsof the vent flow or pressurewavesfrom the vent bursting to explain
the flame accelerationupstream of the vent. There is no mechanism by which
turbulencecreatedby flow throughthe vent can be convectedupstreamand hencethis
 171

mechanismof flame accelerationin vented explosions has always been a dubious

explanationof why the laminar burning velocity has to be increasedby a turbulence

factor of the order of 4 or 5 to give agreementbetweentheory and measurements
(BradleyandMitcheson,1978b).

60 1r-, Cortrelgrition
C2ýVaJr,
04-4, Erd Ignfion 7-z: Pv.
1!ýýPv
BBfft3w
16 :
so I- FV. 175ntw
ý Pv 178ntwl
FV . 209 rrbW 14 Pv: 209ntar
40 PV - 424 mber 424ftwi
12 _'',
'ýPv
lo
30

,
rL 20-, 6-
4,
10 A
2-
oo------ 0
O.D al a2 063 a4 0.5 0.6 0.7 0 0.05 al Ql 5 0.2 0.25 0.3 0.35 0.4

D"=fmmlhawark(rrO Cistarcefromthespark(ni
so I-po. ombff Wair.Erdloi6an
70 1-
Pv - 98 rvtw1 40 - 7-_pv: - WJair, Cove igritan
Orrim
ý-Pv- 178 =
60 -PV eafft)W
-N-208: 35
_P4, . ft. i-Pv . 178rrtw
50
("IA 68) 30 -Pv-2DDnter
ý-PV-424ntw!
,ii-W 25 11, Fe W 613
30 20

20

W 10

0 5
0 111 062 03 114 M5 Ole 0.7 0.8 0
DIOArtefromto spark(ni 0.2 03 0.4 0.5 O's
Dstw, wfmm*mopwk(rrj

Figure 5.16 Influenceof P, on flame speedupstreamof the vent for ethylene/air((D

0.8) andhydrogen/air((D= 0.54) with end andcentralignition.

The results of the maximum flame speedsin the vessel are shown as a function of
equivalenceratio in Fig. 5.17 - 5.20 for methane,propane,ethylene and hydrogen/air
with end andcentralignition. Thesefiguresreport the averageflame speedsmeasuredin
TV2andTV4in Fig,2.3).Theseflamespeeds
thesecondhalf of themainvessel(between
are considerablygreaterthan that for the laminar sphericalflame which 3m/s for
methane,3.6m/sfor propane,5.5m/sfor ethyleneand19.7m/sfor hydrogen/air.

For methane/airprofile, the flame speedincreaseswith respectwith the increasein P,

but the casewhere flame speedwith open vent always gives lower flame speedis not
featuredin propane/airand those reactive mixtures i. e. ethylene/airand hydrogen/air.
For propanelairwith end ignition, Fig. 5.18 shows that flame speed,just upstreamof
 172

entry into the vent duct, was 19 m/s for (D= 0.8 with no vent cover and increasedto 50
m./s for P, = 424 mbar. The laminarburning velocity at (D= 0.8 is approximately0.35
m/s and the adiabaticsphericalflame speedapproximately3 m/s. The measuredname
speedsare muchhigher thanthis andthe increaseis consideredto be due to flame self-
acceleration resulted from the formation of cellular flames
(Kasmani, Willacy, Phylaktou and Andrews, 2007). The maximum reactivity of
propane/airmixturesis at (D= 1.1,asthe peakflame temperatureoccurshere.

CH4/adr
Go
Centr.1 igNtion
50ýý 0 Pv=O
, -m- Pv = 98 nta
40- & Pv = 178 ntEu
x Pv = 209 rnbat
30 w Pv- 424 nba
1
201
gý
lz 10
0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
EqLiWerr,
e atio
40-, CH4/ajr
36 Endlgribon
32

24, i
20-1
161 0 Pv- 0

12 " Pv- 9B n
1& Pv - 178
8
), Pv- Mg
I
)K Pv. 424
0 L-- -.- - -
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Eclivalerre rclio

FigureS.17 Methane/airfor endandcentralignition asa function of equivalenceratio

The resultsin Fig. 5.18 showthat the maximumflame speedwith no vent cover was 52
m/s, considerablyhigher than the 3.6 m/s for a sphericaladiabaticlarninar flame speed.
This increasedto 120 m/s with the 424 mbar vent burst pressure.However, for richer
 173

mixtureswith (D= 1.35the flame speedincreasedto 234 m/s for the highestvent burst
pressureof 424 mbar. For no vent burstpressurethis mixture had a lower flame speed
than for (D = 1.1, as expectedon the basis of laminar flame speeds.For very rich
mixtures of (D = 1.65, which is closeto the rich flammability limit, the flame speeds
reducesharplyto valuescloseto thoseat (D= 0.8. Theseflamesspeedswere20 - 40 rn/s
andarevery fast for sucha nearlimit mixtures.

For the conditionswith no vent burst pressurethe resultsare reasonablyexplainedby

self accelerationof the flame as beendiscussedearlier, which are a function of Lewis
number and Markstein no. Literature showsthat for these are <I for rich propane
mixture, which gives high flame acceleration(Bradley, 1999,Clarke, 2002, Markstein,
1964).However,the effect of the vent burst pressureon the flame speedis difficult to
explain. It is known that increasingthe pressure,which occurs when P, is increased,
acceleratesthe cellular flame effect and makesit occur at shorter distances(Bradley,
1999). This combined with the Lewis number effect, which increases the flame
accelerationfor rich mixtures (Clarke, 2002, Markstein, 1964), are the most likely
causesthe flame speedvariationswith equivalenceratio. It is shown below that these
effectsdirectly causethe peakoverpressureto be greatestfor the (D= 1.35rich mixture.*
Similar resultswerefound for centralignition of propane/airmixture (Fig. 5.18), but the
flame speedswere lower at all P, due to the shorterdistancefor flame self acceleration
(Kasmani,Willacy, Phylaktou and Andrews, 2007). The maximum flame speedstill
occurred at (D = 1.34 with 145 m/s comparedwith 234 m/s for end ignition. For
methane/airmixturesthe resultswerealsosimilar, but the peakflame speedswere lower
and the maximum flame speedwas in the near stoichiometricregion and not for very
rich mixtures,as found for propane.The reasonfor thesedifferenceswith methanewas
the much lower tendency for self accelerationvia the cellular flame mechanism
(Bradley, 1999,Clarke, 2002,Markstein,1964).
 174

240 -j Q3Ft/air

ýýigribon
200,
-70-LPV-0-
-a Pv=98mbar
160 Pv-178mbar
x Pv=209mbar,
120 w Pv-424mbar

80

40-1

0.6 0.8 1.2 1.4 1.6 1.8

EqLi\rdence rzbo

2W -IC, 3F6/ajr

Certreiarition
Pv- 0
160
Pv - 98 ntar
ý6 Pv- 178 rrbar!
120, ý
x Pv- 209 nbwý
w Pv - 424 atmý
80

40-,

m r: f-fEEE
0
0.6 0.8 1 1.2 1.4 1.6 1.8
Ec;ii\derr-a rdio

Figure5.18Propane/airfor end andcentralignition asa function of equivalenceratio

However,sinceonly lean limit can be doneon ethylene/airandhydrogen/airmixtures in

this work due to the safety procedure,it can be said that ethylene/airhas the same
featuresaswhat beenillustrated in propane/airexplosionat lean limit mixtures.With no

vent cover,the maximumflame speedwas 17 m/s and this increasedto 24 m/s with the
424 mbar vent burstpressureas picturedin Fig. 5.19.There is not that much significant
increasein flame speedin various P, for end and central ignition as the occurrenceof
flame cellularity is not great at this lean concentration stage but the turbulence
generatedby the backflow and flame front interactionwith pressurewave substantially
increasethe P.. as shownin Fig. 5.23 later.
 175

30 Q2H14/ü r

Erd igribon
0 Pv -0 nter
1 la Pv - 9B ntar
20
Pv. 178 rrt)arý,
>
x Pv - 209 ntmr'
15 i
.9 NE Pv = 424 ntari
10

0,
0.55 0.6 0.65 0.7 0.75 0.8 0.85
EqLivAerr,e rdio
CýH4/edr
25-1
Certreigriton

20 s Pv= 0 ftar
Pvv
- 98 ntw
15-, 6 Pv= 178nteri
N= 209 ntarý

w Pv - 424 rrbTl
10-1

0, IrII11

0.55 0.6 0.65 0.7 0.75 0.8 0.85

dencer1lo
Ecpj\.

Figure5.19Ethylene/airfor end and centreignition asa function of equivalenceratio

For hydrogen/airexplosion,end ignition showsthat there was not much difference on

flame speedat equivalenceratio for variousP, exceptfor P, of 98 mbar, Openvent gave
sameeffect on flame speedpropagationas highestPvi. e. 424 mbar. The resultscan be
then suggestedthat cellular instabilities developingon the flame surface(due to Taylor
type, auto turbulenceas observedin Kumar et al (Kumar,Dewit and Greig, 1989)) and
longer distancefor flame travel are the importantfactors causingthe high flame speed
inside the vesselregardlessof the Pv effect for end ignition. Meanwhile, this trend is
not featuredin centralignition wherethe flame speedwas 39 m/s at P, = 424 mbar and
23 m/s at initially open venting.This is about2 times higher in flame speedfor high P,
in respect to open vent. Here, it can be postulatedthat self-induced flame is the
 176

the high flame speed insidethe vesselas there is

responsiblemechanismthat triggering
considerableflame had
acceleration taken place before the vent bursts.

45 hVair,

4o Erd igrffion
351
Mý E3 Pv- es n-tar ý
30-
16 Pv= 178 rrlmrýi
25 4m Pv= 209 rnbar
A 20 Pv= 424

15

10-
[Z e
5-1'

01A, I-

a25 U 0.35 0.4 0.45 0.5 0.55 0.6

EqLd\derr,e r.-dio

45
F61aj
r
40-1Certreigrit on
35 Pv= 0 mtxr

30 Pv - 98 nbar
Pv: 178 ntal
25 I&
Pv 209 nt w
20 Pv= 424
15

10-
i
5ýI-

OL3 0.35 0.4 a45 0.5 0.55 0.6

EcFAWerreredio

Figure 5.20Hydrogen/airfor end andcentreignition asa function of equivalenceratio.

5.1.5 The duct flame speedsand gas velocities

The peakmeasuredflame speedsprior to the flame enteringthe duct pipe are shown in
Fig. 5.9,5,10,5.15-5.16,5.17-5.20as well as in Table 5.1. The associatedUnburntgas
velocity and the turbulent burning velocity upstreamof the vent, basedon this flame
speedand adiabaticburnt gas expansionare also shown in Table 5,1. The turbulence
burning velocitiesmeasuredwith central ignition give a ST/SLburning velocity ratio of
 177

approximately 5, similar to that used by Bradley and Mitcheson (Bradley and

Mitcheson,1978b),However,asshownabove,the mechanismof the flame acceleration
is not due to flow turbulence.Simple momentumconservationallows the unburnt gas

velocity in the ductjust prior to the flame entry to be predictedand theseare also shown
in Table 5.1. This calculationassumesthat the flame behavesas a piston, which is not
the reality as the flame is 'pear 'shaped and not flat. However, the resultant gas
velocitiesare very high and greaterthan sonic in a few cases(incompressibleflow was
assumedin the momentumequation).This indicatesthat nearsonic flow is likely in the
vent and the absolutepressureratio, P2/P6,acrossthe vent duct in Fig. S.II to 5.14
confirmedthis observation.It is importantto realisethat this duct flow is driven by the
fast flamesupstreamof the vent,shownin Fig. 5.9,5,10,5.15 -5.16 and 5.17-5.20.

The flame speedsin the vent duct were measuredand tabulatedin Table 5.1. These
were very fast flames even for an open vent. They were consistentlyhigher for end
ignition due to the fasterflames and the predictedfaster unbumt gas velocities in the
duct. The flame speedsare reasonablyclose to the predictedunburnt gas velocities
computed from the upstreamflame speeds.There are generally higher due to the
influenceof theturbulentburning velocity. However,it is clearthat the dominantcause

of the fast flames in the vent duct is the fast upstreamflames detailed above. The
unburnt gas velocity aheadof the flame in the duct has beenestimatedas 80 % of the
flame speedin Table5.1. For an adiabaticexplosionthe gasvelocity aheadof the flame
is (E-I)/E timesthe flame speedsandthis is 87 % of the flame speedfor stoichiometric
hydrocarbons.The lower value has been used to account for duct wall heat losses
(Phylaktou,FoleyandAndrews,1993).

In the caseof equivalenceeffect, the high flame speedupstreamof the vent induces
higher unburnt gas velocities ahead of the flame. These high vent pipe unburnt
velocitiesresultsin a very high pipe turbulencelevel and consequentlyvery high flame
speedin the duct pipe as shown of the worst case of propane/airin Fig. 5.21 and
hydrogen/airin Fig. 5.22. This showsa very large increasein flame speedinside the
vent duct with P, from 200 m/s to 700 m/s as the P, increasesfrom zero to 424 mbar in
the caseof propane/airat endignition for (D= 1.0.In the caseof hydrogen/air,the name
speedinside the vent duct was 300 m/s for P, =0 and increasessharply to 550 M/s for
Pv = 424 mbar for (D = 0.54. The causeof this is the large increase in flame speed
 178

upstream of the vent in Fig. 5.20 from 23 to 40 m/s corresponding to the increase in Pv-
This high flame speed inside the duct pipe at lean concentration in hydrogen is

responsible to induce high turbulent flow leading to an intense mixing of cold and hot

gaseswhich bringsto a subsequent

violent combustion(bum-up)inside the duct during
re-ignition (Ponizy and Leyer, 1999a,Ponizy and Leyer, 1999b).Bum-up in the duct
has then to be considereda crucial phenomenaaffecting the final overpressureof the
vesselandin this case,to leadto a detonation,

wo , CArair
1 End ignition
7001
s Pv-o
Iz J
600 la Pv- 98 mbar
i tj Pv-. 178 barý
500
Pv=209mbari
m
4W - PV- 424 mbar

300

200 J'

100

0
0.6 0.8 1 1.2 1.4 1.6 1.8
Eclim3lerce rdo

Figure 5.21 Flame speedin the duct pipe for propane/airas a function of equivalence

ratio at endignition,
 179

600 ] Wadr
! Certre lgribon
500-,
. A Flv ntw
=0
Pv- 9B ntar
4W - Pv- 178 rrbarý
x Pvý 209 rd",
300-, Pv- 424 n-bari
,
.9
200-1,

Ir 100 -,

0
0.25 0.3 0.35 0.4 0.45 o.5 0.55 0.6
EqLi\rcderr,
e rc-dio
Figure 5.22 Flamespeedin the duct pipe for hydrogen/airas a function of equivalence
ratio at endignition.

5.2 The influence of equivalenceratio on Pmax

P.. is shown as a function of the equivalenceratio in Fig. 5.23-5.26 for methane,

propane,ethyleneand hydrogen explosionswith end and central ignition. For an open
vent, Fig. 5.24 shows that P,,,. was higher than with P, = 98 mbar for propane/airwith
end ignition from (D = 0.8 to 1.12.The same figure also showsthat for central ignition
for P, = 98 mbar and 178 mbar from (D=
the openvent had much higher overpressures
0.8-1.15.This did not occur in the methane/airexplosionsin Fig. 5.23, wherethe open

vent was alwaysthe lowest P.. for all equivalenceratios. Theselower overpressures
for 98 mbar and 178 mbar vent burst pressurescomparedwith open vents for the
propaneexplosionsare difficult to explain. The upstreamflame speedsresult in Table
5.1 doesshowinga higher flame speedfor the openvent for (D= 1.0 and at (D= 1.37 as
shown in Fig. 5.18. Propaneis propensityfor cellularity at rich concentration(Tseng,
Ismail and Faeth, 1993) and this would explain the higher P. in comparison of
"
methane/air.For propane,the onsetof cellular flamesoccursover a shorterdistanceand
the self accelerationof the flame is greater,which is a function of the distancefrom the
spark.The significant effect from this is to acceleratethe flame towardsthe vent more
quickly than for methane/airexplosions,as shownin Fig. 5.15. The effect is to stretch
the propaneflame so that it is more elongatedthan a methane/airflame. The action of
 180

the staticburstpressureof the vent is, for low burstpressures,to delay the acceleration
of the flameto the vent. The flame doesnot know that the vesselis venteduntil the vent
burstsandwaspropagatingasin a closedvessel.For low ventburstpressurethis givesa
smallerflame areawhen the vent burstsandit is the effect of the vent burst pressureon
the flame bulk shapeand not on its centreline velocity that causesthe pressureto be
reduced. It is consideredthat this effect hasa greaterrelative influence on the propane
flamesdueto their greateraccelerationto the vent andthis is why propaneandmethane
flamesbehaveddifferently for low vent openingpressures.A higher initial flame speed
would indicatea higher massbum rate and hencea higher massflow rate in the vent
pipe and this would createmore turbulenceand hencehigher overpressuresdue to the
fasterflame speedsin the vent pipe.

Unfortunately,this was the only result that gavea higher flame speedprior to the vent
and hencethis mechanismcannotexplain all the results.Further,it is very significant
that only the propaneexplosionsdemonstrated
this effect. This is also indicatedthat the
quickeronsetof cellular flamesandthe greaterself accelerationof propaneflames may
be the causeof this difference.It is conjecturedthat P.,,,, was generatedby the rapid
turbulentcombustionof the unburntgasesleft upstreamof the vent in the main vessel.
The flame preferentially acceleratesin the direction of the vent once this is open,
leaving a large proportion of the unburnt mixture in the ventedexplosion vessel.The

gasesin the outerpart of the vesselare unbumtwhenthe flame entersthe vent. A high
flow back pressureis generatedby the high velocity unbumt g4sflow in the vent and
high flame speedsin the vent pipe, asshownin Fig. 5.21. This forcesthe
the subsequent
upstreamflow to reverse and ignite the outer unbumt gas flow in a fast burning
turbulent flame. This then acceleratesthe burnt gas flow out the vent pipe, which
increasingthe pressurerise.
 181

CH,/air
CH4/air Co*e Igrition
Eridigriban 3.0 j-. *-Pv-ontEr
2151!
-6- Pv. 0 1
!
--o- FV. 98 fftw
- Pv. 96ntx
ao ý 2.5 -a- Pv .M rrbar,
Pv. 178 ntEr,
-G-
-. m- Pv - 209 rrberý
-, m- Pv- 209 ntrr
IT
20-,,
i --w-Pv»424 or,
x-
2.0-1

ts
or

a5i
;

ao 0.0
116 0.7 018 a9 1.1 1.2 1.3 1.4 0.6 (17 0.8 OL9 1.0 1.1 1.2 1.3 IA
Eco%derre refio Ectkderce rgdio

Figure5.23Methane/airat end(left) andcentre,(right) ignition at variousP,

7.0 - C"air
End lgribon SO QkWir
4.5 Certrelgrit
-ä- Pv: 9Bn F'-Oýl 6
5.0- 4.0.
-a.- Pv 178 ý--A-PV-98
as
4.0 - -'»- - MB Pv- 171
*-PV. 4m ao- PV-2(X
Z5 - pv.
10-
1.5
to
US
0.0 0.0"
0.6 018 1 12 1.4 1.6 1.8 as 0.8 1.2 1.4 1.6 1.8
EcFJ%egerce
reMo EcpjKa6, ce rido

Figure5.24Propane/airfor end (left) andcentre(right) ignition at variousP,

However,for higher vent burst pressures> 200mbar,P..,, always increasedrelative to

the open vent casefor both methane/airand propane/air.This was due to the greater
flame area and mass burning rate when the vent opened,together with the sudden
accelerationof the gasesin the vent pipe when the vent burst due to the imposed
pressuredifference.This produceda large increasein the massflow through the vent
and hencein the overpressure.It is consideredthat this significant influence of cellular
flames with propaneexplosionswill be greaterfor ethyleneand hydrogen explosions

andthis is shownin Fig. 5.25 and5.26 for ethylene/airandhydrogen/airrespectively.

 192

C,2Hlä r IZO C"aly

ELO '
i Erdigribon Certrelgibon
ý
7.0 v-0 nt et
j
ýý-*-N«sBntw
ýPv: 178ntm
ý-a- PV- 178ntorý "o-,
- -*-Pv MDrrte

4.0

20

ao
(10 --I-- ----
OL85
a5o 0,55 OAD 0.65 0.70 OL75 a8o
0.50 0.55 (160 a65 VO CL75 (180 US
Ecdv,4erce rabo
EqLi\oWerre
mk

Figure5.25Ethylene/airfor end (left) andcentre(right) ignition at variousP,

wat
45 kvalr 120 C"eigtifion
4.0 Erdignton
IýFV-Offbar Pv-Ontff
aI51
ýPV. 98rfbar Fv - OBrrtxr
3.0 1-*- Pv- 178Ma' - Pv. 178nbar,
Pv- 209ntivi 8.0
iý- Pw: 209ntwý
'Lý 2.5- FV- 424ft&
Pv 424nbý
201,

4.0
1.01

20
ao
a3D (135 0.40 a46 aso OM a60 0.0 i 10,
--
E4mmcerado (130 0.35 0.40 0.45 0.50 0.55 0.60
EcovJemsrado

Figure5.26Hydrogen/airat end (left) andcentre,(right) ignition at variousP,

-

For both gas/mixtures, ignition

central gavehigher P.,,,,as a functionof equivalence
to thetwo hydrocarbon/air
ratio compared mixturesinvestigated. As tabulatedin Table
5.1,thepeakP.. is theresultantof detonationspikeoccurredin the pressure
tracesfor
bothmixturesandit wasabovethemaximumadiabaticpressure insidetheclosedvessel
of 8 barsat (D= 0.8for ethylene/air
and(D= 0.54for hydrogenat centrallyignited.This
observationbringsout an importantfact that high burningvelocitymixturessuchas
hydrogen/airandethylene/airmixturesbehavedifferentlyfrom thosewith low burning
andformationof flamecellularityis anotherfactorthat
velocities.Theself acceleration
enhances themassburningrateinsidethevesselbut theconsiderable amountof unburnt
gasesleft trappedinsidethevessel(mixturereactivity)is theimportantfactorthat leads
to theincreasetheP,,,,,,
with detonation
spikeobserved.
 183

5.3 Comparisonwith other experimental data

In order to developunderstandingon the phenomenologyof duct ventedexplosion,the

current work resultswere comparedto other published experimentaldata (Russoand
Benedetto, 2007, McCann, Thomas and Edwards, 1985, Molkov, Baratov and
Korolchenko,1993,Ponizy andLeyer, 1999a,Ponizy and Leyer, 1999b),Later, all the
experimentalworks including the currentproject will be comparedwith the correlation
offered by NFPA 68 to investigatethe discrepancybetweentheoreticalapproachand
experiments.

Figure 5.27 showed the Predwith duct as a function of the reduced pressure obtained in
the same tests without the presence of duct. On the same figure, all experimental data

were compared to the correlation given in NFPA 68 for duct pipe length in between 0 to
3 m. As expected,the correlation given by NFPA 68 is under predicted the experimental
data results, suggestingthat there is more complex interaction between the vessels with

the presenceof duct with respect to the ductless vessel explosion i. e. simply vented. It is

also interesting to note that even though the same reduced pressure obtained in vented
explosion without the duct pipe attached, there was a variance Of Predin the presence of
duct pipe. This situation implied the correlation associated P,,d with the duct pipe and

without the duct pipe should not be taken as simple as given in NFPA 68 but other
parameters should be taken into account to accommodate the increase Of Predwhen the
duct pipe is attachedwith respect to simply vented explosion. As previously discussed,
ignition position, vessel volume, self-acceleration, induced turbulent, pressure loss as

well as secondaryexplosion should be included in the correlation in order to fit with the
results obtained experimentally. All published experimental data are listed in Appendix
A (Table A. 2).
 184

10.0 -1
9*01 DeGoW and ChEtadit

8.0 t) Kbkovet d
c
7.0 ,& KtCarn et Ed
6.0- x Wharddr
x Prcpardar
5.0-,
o EthylenfMr
"": 4.0 0
+
3.0 NFPA68
2.0 - Badknoctt'se4jWicn,
1.0 ý
0.0 )r-I
0.0 20 4.0 6.0 8.0 10.0
ckict(bar)
PiedWthoLA

Figure 5.27 Comparisonbetweenotherpublishedexperimentaldataand currentproject.

The datareportedwere 10 % methane/air,4.5% propane/air,5.2% ethylene/airand 16
% hydrogen/air and 5% acetone/air.NFPA 68 correlation (NFPA68,2002) and
Bartknecht'sequation(Bartknecht,1993)are usedfor 0< Ld >3m.

Another interestingpoint to highlight is the influenceof IJD to Pw. It is clearly shown

that LID is not main factor influencingthe increasein Pttdon the presenceof the duct
pipe.As shownin Fig.5.28,Predwasalmostinsignificantat the different L/D obtainedin
Ponizy and Leyer's work (Ponizy and Leyer, 1999a,Ponizy and Leyer, 1999b)at end
ignition. The same LID gave insignificant P,m at the different mixture reactivity
(proparie/airand methane/airmixtures)illustrated by DeGood and Chartrathi (DeGood
and Chartrathi,1991),McCannet al (McCann,ThomasandEdwards, 1985)and results
from this works either end or centralignition. This condition can be suggestedthat LM
is not significantin influencingPredwhenthe duct pipe attachedin low burning velocity
mixturesbut not in high burningvelocity mixturesi. e. hydrogen.
 185

0 Pbrizy and Loya ( Endigritiort propaiWair)

a DeGOWaid ChEd rah ( PropanEVair)
9.0 A Molkov el A (AcEfonwair)

* MoCarnal al ( MEtheriEVair)

ao * Pt)nzy " Loyu (Calreigntiork propmeVair)

* Methwi0air (Erd igritiori
7.0 * MothEnYair (CA3rtroigntiorV
Cl * RopaneVair (End igritioro
6.0 * Prop2mar ( Centreogrilico
A " FWrogaVair ( End gdlioro
so + " ýJodrogcrYair (CA31reigritiorV
+ Bt~ar ( End igFition)
g 40 ý A BhýkrWair (CAylreigritior#
)e
C- 4
30 &0 *
ox
00 0
0000
20 x
m-
1.0 -r1
x

0.0
0 50 100 150 2DO

UD
L

Figure 5.28 Predas a function of LJD

5.4 Concluding remarks

The high overpressure due to the addition of a vent duct, to a 0.2 m3 closed vessel with

an LJD 2 K
of and of 16.4, occurs after the flame has exited the vent duct, but is not due

to an external explosion. The main effect of the ignition position is to give a greater
distance from the spark to the vent and greater flame acceleration. End ignition gave
higher Pmaxfor low burning velocities mixtures but central ignition is the worst case

scenario in Pnaxfor hydrogen/air and ethylene/air. It was found that substantial amount

of unburnt gases left inside the vessel after the vent burst is the leading factor in

increase of Prnaxfor high burning velocity mixtures for centrally ignited. The associate

gas velocities ahead of the flame create high unburnt gas flows conditions at entry to the

vent and this give rise to high back pressures.

As P, increases, the distance of normal spherical flame propagation increases and there
is a further reduction in acceleration distance. This initially reduces the overpressure at
low P, The effect of the vent burst pressure is to increase the flow velocity in the duct

when the vent burst, as the flame has had more time to grow upstream of the vent and
this gives a higher vent duct flow velocity once the vent cover bursts. The effect of the

vent burst pressure is complex and non-linear and is not represented by the linear effect
in the correlation of Bartknecht (Bartknecht, 1993) used in NFPA 68 (NFPA68,2002).
 186

The greaterself accelerationof flames due to cellular flame developmentfor propane

relative to methaneis shown in the results.Propanehas a very strong influence of
equivalenceratio on the flame speedsand overpressures,especiallyfor rich mixtures
wherethe highestflame speedsand overpressuresare exhibitedfor 4) = 1.35. It cannot
be assumedthat the mixture with the highestlaminar burning velocity, measuredon
small diameterflames with no cellular flames, is the mixture with the worst case
explosion hazards,For gases such as propane that have a strong cellular flame
developmentthe worst caseexplosionhazardwill be for rich mixtures,wherethis effect
is maxin-dzed.In the presentwork this was (D = 1.35 for propane.For methanethe
ventedoverpressureoccurredat the samemixture strengthas that for the maximum
laminar burning velocity, but the cellular flames did contribute to the higher flame
speedsand overpressures. This is becausefor methanerich mixtures do not have
greatercellularity.At leanconcentration(low equivalenceratio), high burning velocities
mixturesexhibit a detonationspike in the pressuretracesinside the vesseleventhough
the behaviourof secondaryexplosion( bum-up)andpressuredropsat the duct entrance
reproducewhat have beenobservedin methane/airand propane/airmixtures well after
the flamehasleft the vessel-ductassembly.

The correlationoffered in NFPA 68 (NFPA68,2002) and from Bartknecht's work

(Bartknecht,1993) for the calculation of Pred with the duct was shown not to be
conservativein comparisonwith the experimentaldatareported.It also highlighted that
L/D is not the importantfactor in increaseOf Predwith the duct pipe in propane/airand
methane/airbut not for high burning velocity mixtures.
 CHAPTER6

VENTED DUCT GAS EXPLOSIONS WITH

BIGGER DIAMETER PIPE ATTACHED
 187

6.0 Introduction

The influence of duct attachedto the main vesselon P,,,,,,has been investigatedin
earlier chapter.The investigation confirms the findings of the previous works and
highlights someconclusivefactors that lead to the severity of vented gas explosions
with the duct pipe attached. From the work, major influences on the behaviour of
duct
ventedexplosionswith attached pipe which leadsto a severePma,,
are;

Increasingthe aerodynamicresistanceto flow as high unbumt gas velocity

aheadof the is in
vent experienced most of the cases,
causing it choked and
reducingthe flow of unbumt gasesout of the vesselaheadof the flame front. It
thenincreasingthe amountof trappedunbumt gasesinside the vesseland hence
the Pinualso increases.The bum-up mechanismor secondaryexplosion inside
the duct pipe causesthe back flow towardsthe vesseland the interaction with
the flame front will generateturbulent pressurewave and thus, enhancingthe
pressureinside the vesseland at some casesin high burning mixtures, it will
lead to detonation.

The acousticoscillation inducedin the duct that generatepressurein the vessel

and thus modify the flame propagation inside the vessel and been studied
extensivelyby Kordylewski and Wach (Kordylewski and Wach, 1988). The

coupling of the flame with acoustics,flame turbulenceinteraction and sonic

blockage in the duct were the possible explanation of the high pressure
inside the vesselfor ventinggasexplosionwith the duct attached.
generated

In order to reducethe peak over pressurein an explosionwith a vent duct to that of a

simply ventedexplosion(ventingwithout the duct pipe attached),the vent areaand vent
pipe diameter needs to be increased (2002), but therehasbeenno specific experimental

validationof this procedureas this topic has been sparselyresearched.In the presence
of a vent duct, an increaseof venting area and duct diameterhas been found to not
alwaysresult in a decreasein the peak over pressure(Ponizy and Leyer, 1999a).This
returnsthe focus to the more realistic scenarioof whetherthe similar behaviourwould
 188

be observedwhenthe bigger duct pipe is attachedin comparisonto the duct areais the
same as the vent.

The main reasonfor the increasein the overpressurewhenlong vent ductsare attached
to vents is due to the phasein the explosionwhen the flame is in the vent pipe with
unburnt gas mixture aheadof it. The expansionof the burnt gases in the vent pipe
the
greatly accelerates unburnt gas flow and this increases
the vent pipe friction, inlet

and exit pressurelosses(Lunn, CrowhurstandHey, 1988).Theseeffects are a function

of the dynamicpressurein the vent pipe. In principle, the dynamicpressuresin the vent
pipe can be reducedby simply usinga largervent diameterthanthat for the vent, rather
than increasingboth the vent and ventpipe size.For example,if the vent pipe wastwice
the diameterof the vent then the vent pipe dynamic pressurewould be reducedby a
factor of 16,if the vent massflow rate is remainedconstant.Someevidencethat a larger
vent pipe diameterwould reducethe overpressurewith no changein the vent size was
provided by Nagy (Nagy and Verakis), which is quoted in NFPA 68 (2002). Nagy
investigatedtheinfluenceof a 35 % increasein the vent duct diameter(83 % increasein
area) for the samevent diameter and comparedthis with an explosion with a 35 %
increase in vent diameter and pipe diameter. Vent pipe lengths up to 5 m. were
investigatedand showedthat the overpressurewherethe vent pipe alone was increased
in diameterwas only 7% higher than that when the vent and pipe diameterwere
increased.From the experimentsperformedon dust explosions-withvent pipes (Lunn,
CrowhurstandHey, 1988),Hey (Hey, 1991)hassuggestedthat the techniqueof using a
larger vent duct diameterthan the vent diameteris effectiveif the duct area/ventareais
about - 2-2.5 and when P.....is less than 0.5 barg. It is consideredthat this approach
would be a simpler method of designing for safe vent pipes and the present work
investigatedfor gasexplosionsa vent pipe that was closeto twice the vent diameter,as

recommendedby Hey (Hey, 1991).

To achieveequivalentventing with a larger vent pipe, a 0.5 m flange openingin the

dump vesselwasattachedto a 0.315 m diametervent pipe which was Im long ( L/D =
3.2) and 1.94timesthe diameterof the 0.162 m diametervent as mentionedin previous
chapter( Test vessel3), The same0.162 m diametergatevalve was attachedto the vent
and dischargedinto the large vent duct, All the ducts were pressurerated at 25 bar, as
detonationin the vent pipe was known to be a possibility with overpressurein the 15
 189

bar rangepossible(fida, KawaguchiandSato,1985,Kordylewskiand Wach, 1988,Lee

and Guirao, 1982, Ponizy and Leyer, 1999a,Ponizy and Leyer, 1999b, Russo and
Benedetto,2007). If the Im long vent pipe was removed than the vent pipe was
reducedto the 0.3m of the gate valve and dump vesselconnectingflange (UD 1.85).
This was effectively a free ventedexplosiondischargeand this configurationwas used
as the baselinefree ventingcondition. The ventedgasesdischarged into a 50 M3 dump

vessel,as shown in Fig. 2.3b. The ventedgases50 m3 containmentdump vesselwas

pressureratedat II barsasa detonationin herecould not occur.

In this presentwork, only uncoveredvent casewas carried out with four different gases
(methane,propane,ethyleneandhydrogen)at different equivalenceratios. All mixtures
were ignited at the end wall and/orat the centreof the vesselcentrelineby an electrical
sparkwhich gives16 J energiesfor the gasexplosiontests.

6.1 Effect of duct diameter on P..

Figure 6.1 shows, for the most reactive methane/air mixture, the pressure-time profile
inside the vessel at P, for the short 0.162m diameter pipe, which is effectively a free
discharge, and for the two Im long vent ducts of 0.162 and 0.315m diameter

respectively. The results show that the larger pipe diameter (duct area / vent area, AýA,

= 3.78) has little effect on the overpressure compared to that with the vent duct the same
diameter as that of the vent. This was not the expected result and did not agree with the

results of Nagy (2002, Nagy and Verakis) or Hey (Hey, 1991). However, this situation
is not been observed in hydrogen/air explosion as illustrated in Fig. 6.2 where vented

explosion with bigger pipe attached (Dp = 0.315m) shown the pressure peak is about 2
times higher than vented explosion with duct diameter of 0.162m. Interestingly, the

peak pressure for vented explosion with 0.162 m diameter pipe attached gave the same
peak pressure as the baseline case Le, simply vented. Interestingly, this trend is not
observed for end ignition where maximum pressure is lower for baseline case in respect
with duct vented explosion as shown in Fig. 6.3. Again, for centre ignition, the pressure
rise inside the vessel is directly influenced by the mixture reactivity left inside the
vessel after the flame exited the vessel and the presence of the duct gave little effect to
the final value of pressurein this case. This observation contradicts with the other works
 190

wherethe presenceof the duct pipe attachedwith the vesselgavehigher P,,,,,

x compared
to simply ventedexplosion.

1.8 CWair, 0=1.08

1.6-, Endigniton
1.4
Smplywnted
I
1.2
1.0 IDA 0.162
2
-Duct%ented,
D 0.8- m, Ld= 1.0m"
0.6 Duct%ented, 15
d, Dd 0.315
U)
m, Ld 1.0) m"
)m.
M.
a) 0.4
0.2-,
0.0
0 0.05 0.1 0.1 0.2
-0.2
-0.4 -' Th-e(s)

Figure 6.1 Pressuretime profile for methane/airat (D= 1.08

FVadr O=0.54
8.01
I Certre igrition
7.0 1,
... Smplywnted
6.0
D_Jct"e nted DP_O. lr "2
5.0 Duct wnted, Dp 0.315 m

4.0

3.01'
ý.o

1.0

0.0
1) 0.02 0.04 0.06 0.08 0.1 0.12 0.14
-1.0
i Time (s)
-4u,

Figure 6.2 Pressuretime profile for hydrogen/airat (D= 0.54

 101

4.0 Wair 0=054

Erd igrition
3.5

3.0 Simplyýented

2.5 ented, Dp=0.162m

-Duct\,

2.0 ý
-Duct \.ented, Dp - 0.315m
1.5
La)

1.0

0.5

0.0
ý. 03 0.05 0.07 0.09 0.11
-0.

T ni--(s)

Fi"Ure 6.3 Pressurethrie profile for hydrogen/air at end ignition. (D = 0.54

From Fig. 6.2 and 6.3, it can be deducedthat ignition position gives important inflUenCe
L-
on the final pressure inside the vessel. As discussed in previous chapters, the
combustion inside the vessel is almost completed when the flame exited into the duct
pipe, leaving only small amount 01'LinbUrntgasesinside the vessel which results M high
flame speedsand hence high unburnt burning velocity prior to duct entrance. Tile final

pressure inside the vessel is directly due to the high turbulent reversal flow from the
duct pipe to the vessel from a violent burn-up or secondary explosion and the flame
front interaction with the pressure/shock waves. In other words, there is a mutual
interaction between the vessel and duct during the explosion development marking to

the final pressure inside the vessel. For central ignition, the interaction between vessel
and the duct seemsnot playing an important role in order to determine the final pressure
hUt dUeto the substantial amount ofunbUrnt gasesleft inside the vessel.

TO FLINherInvestigate the overall trend of' vented gas explosion with bigger diameter

pipe attached, as a FunctiOll 01' e(lLIIVaIeIICCratio I*or all StUdied gases were
presented in Fig 6.4 - 6.7. For mellialle, at Q) = 0.84, the larger vent duct had only a
slightly larger overpressure than J'Orthe near 1'reevent condition and at (1)= 0.68, the
overpressures were the same. However, for tile 0.162 in Vent (ILICtthe OVerpreSS111-CS
were always much higher than for tile freC Vent it 111eLjLIiVIleIICCratios but slightly
lower in respect with larger vent p1peat (1)= 1.05 for end ignition. The saine trend is
 192

shown in propane/airwhere the rich mixture concentrationgave high P., for larger
vent duct. However,in ethylene/airexplosion,it seemsfor the attached0.162 m vent
were higher than for the duct with bigger diameterattachedas
duct, the overpressures
shown in Fig. 6.6. From this data, it can be deducedthat it was only for the most
reactive mixtures that the larger vent duct did not solve the problem of the large
increasein the venting overpressurewhen the duct was the samediameteras the vent
andthereasonsfor this wereinvestigatedfurther.

CH4/adr
1.6 D-ctvertecL10 -
(1315rnendIgrition
D-xi vwtecL Dp
1.4 0.315MpentrEdotionon

1.2 Dict vertecLDp -

1&0.162 m erd igrifion
1.01
- 0 Dict verted,
0.8 Cp.0.162rricentral
Vrrýyverted,
0 end
0.6 bMion
Srrplyverted,cerdral
0.4-,
kWion
0.2 j;
0.0-

0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2

Eqdvalercerat o

Figure6.4Pmax
of methane/airasa function of equivalenceratio
Sn'PiyveMed,endlgrülon
a5 -9 C.3hWair
EI Srrptyverged,certre igrition

i& Dicivefted, P>ü162rnend 12

215- kmkxl
--0-- Mict vergEct Dp- 0,162m certre
Vtlan

W D-U verted, C4>«315Mend

Vtion
*E Dxt vertad, CP- (1315mcortre
,0
1.0 1

0.5
0.0

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

EqLfi\dence rello

Figure6.5P.. of propane/airas a function of equivalenceratio

 193

The overpressures
in Fig. 6.4 and 6.5 for the mostreactivemixture (P,,,,,,= 1.49bar for
(D = 1.08for methane/airat end ignition and P,,,,,,= 3.26 bar for (D = 1.375at centre
ignition for propane/air)cannot be predicted from recommendedcorrelations by
Bartknecht(Bartknecht,1993)as written in Eq. 5.3 and5.4. As mentionedin Chapter3
for simply ventedexplosion,Bartknecht(Bartknecht,1993)gavethe correlationfor 100
mbarstaticventburstpressureas;

1=0.1265 log KG - 0,0567

K Prod 0.5817
(6.1)

The free vent overpressure is predicted by Eq. 6.1 to be 5.45 bar for methane/air

comparedwith the measuredvalue of 0.35 bar. It is 7.46 bar for propane/airfrom Eq.
6.1 and 1.62 bar experimentally.It is clear that Eq. 6.1 cannot be applied to smaller
vesselvolumes.Examinationof the venting dataof Bartknecht(2002) showsthat all of
his vesselvolumes had lower overpressuresthan for Eq. 6.1 and that this was the
correlationfor his datafor a 10 m3vessel.If the measuredfree vent overpressureof 0.35
bar for methaneis takenthen the effect of a vent duct < 3m long is predictedby Eq. 5.1
and 5.3 to be an increasein the overpressureto 1.23bar and only 0.5 bar respectively,
well below that measuredexperimentally.However,if the correlationin Eq. 5.4 for duct
length> 3m is usedthen the predictedincreasein the measuredoverpressureis 1.44bar
for methaneand3.18 bar for propane,which is closerto the measuredresultsin Fig. 6.4

and 6.5 for methane/airand propane/airrespectively.Yet, Eq. 5.2 given for duct length
in between3m and 6m seemsto give under predictedvalue of 1.03 bar for methane

and 1.44bar for propaneif comparedto Eq. 5.4. Summaryof the predictedvaluesbased
on Eq. 5.1-5.4andexperimentalresultsis shownin Table6.1.
 194

o Smplywented,endigrýbon C,2Hýai r
9. 1
la 9mpdy\ented,contreigniton
8,0 -ý
7.0-1 6 Duct\ented, Dp-0.162m,
; end ignifion
EC) -ý
ý --o--Ductm. -nted, Op-o. 162rn,
5.0- centre ignißon
40 --X- DUC manted,Dp - 0.315 m,
-'
erd ignifion
NE Duct \ented, Dp=0.315m,
20 cantre ignibon
7
1.0-",

0.0

()0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85

_J.
EqLi\dence räio

Figure 6.6 P.,,,.of ethylene/airasa function of equivalenceratio

Table 6.1 Summaryof experimentalresults(highestP.. eitherend or central ignition)

with predictedvaluesfrom Eq.5.1,5.2,5.3 and 5.4 where Predwithout duct obtained
from the simplyventedexperimentaldata.
Gas/air Expt Expt Eq.6.1 Eq.5.1 Eq.5.2 Eq.5.3 Eq.5.4
data data for
with simply
0.315 vented
m dia.
duct
Methane/air 1.49 0.34 5.44 1.23 1.03 0.49 1.42
((D=1.08)
Propane/air 3.26 1.62 7.46 2.37 1.45 1.87 3.18
((D= 1,37)
Ethylene/air 3.82 0.50 10.93 1.36 1.06 0.68 1.73
((D= 0.8)
Hydrogen/air 6.50 4.78 14.59 5.80 4.57 4.77 5.56
((D= 0,54)
 195

that Eq.5.4 for duct lengthin between3 and 6

From Table 6.1, it can be recommended
m derived by Bartknecht (Bartknecht, 1993) gave reasonablevalue compared to
experimental data.As discussedpreviously, 5.4
Eq. appliesfor sonicventing conditions
as well as long vent pipesand it will be shownbelow that the venting conditions are
to
close sonicwith a vent pipe attached. It can be postulatedthat the presentresultsare
quite at variancewith the correlationsfor vent designand vent ductsin Eq. 6.1 and 5.4
respectivelyand further work is recommendedin the reliability of these correlations,
for
especially smallervolumesand high K. It is possible that for smaller K the lower

vent flow velocitieswould createless turbulencein the vent ductsand the presentlarge
back pressureswould be eliminated.The effect is essentiallythe sameas using leaner
nlixtures with the At
presentcase. lower K the velocities in the vent pipe would not be

ashigh for the mostreactivemixture andthe backpressureincreasewith the vent would

be lower.

7.0 1o Snioy %otecýendlgrition He/air

ý
ao --4a-GnI3lyeritect cenre
lgrition
SO, Duct \Wecý Dp-0.162n
end igriticn
4.0-4 OW %Me Dp-0.162rr%
cergreigritlan
ao )E Dict writecý Dp. 0.315ril
end igrition
2-0 j1
ME D-Et NMe DP,431. IM
caire Igrition
1.0

0.0 ! fvw-'-
(130 0.35 0.40 0.45 a5o 0.55 0.60
Ec;Av.-derr,e redio

Figure6.7 P..,, of hydrogen/airasa function of equivalenceratio.

However,the trend showedan impingementin the caseof H2/air from previous

experimental in
resultsobtained At
hydrocarbon/air. lower magnitudeof
concentrations
P,,,. is higher with larger duct pipe comparedto 0.162 m diameter duct pipe with
respect to hydrocarbon/air's profile as shown in Fig. 6.7. Surprisingly, higher
overpressureis observedfor simply ventedresult at (D = 0.54 for central ignition in
comparisonwith vented explosion with 0.162 m diameter duct pipe attached. As
illustratedin Fig. 6.3, a suddendetonationspike appearedin all cases,venting with the
duct pipe attachedand simply vented.This peculiar result envisagesthat the violent of
 196

bum-up mechanism inside the duct can be ruled out to be the important role for the final

overpressure inside the vessel in the case of centre ignition. Conclusion can be drawn
based on this observation; the significant rapid rise of final pressure inside the vessel or
'detonation spike' depends on the flame instabilities (sudden venting can give rise to
flame instabilities and rarefaction wave (Dorofeev, Bezmelnitsin and Sidorov, 1995)),

auto-ignition/auto turbulence of the unburned pockets of mixtures inside the vessel

which increase the burning rate due to faster flame and the scale of flow distortion. The
plots of flame time arrival at the comer of the vessel at the spark end where a flame
arrival thermocouple was located would be best illustrated for auto-ignition condition
observed in the present case (refer to Fig. 6.8). If the plot is also valid for central
ignition, then it is confirmed that auto-ignition is the significant role for the high

pressure rise or 'spike detonation' observed in the hydrogen/air tests. The details of
deflagration-to-detonation mechanism are also discussed in Chapter 4.2 for simply

vented case. It then can be postulated that with central ignition, the bigger quantity of
fresh mixtures left for the residual combustion in the vessel is the main contribution of

the high magnitude of 'spike detonation' pressure in comparison with the pressure peak

attained at end ignition inside the vessel. However, as observed before, 'detonation
spiky pressure' is not exhibited in methane/air as methane does not auto-ignite easily
but it can undergo highly turbulence combustion.

Fi2/,-air
1.2
End igribon * Time of spike (simply%ented)

* Time of spike (Dp 0.162m)

1.0 =
* Timeofspike(Dp=0315m)

* Massel comer (simplywrited)

0.8
6 Messeloorner(Dp=0.162m)

0.6 6 \A3sseloDmer(Dp=0,315m)

'o 0.4
X--

0.2

0.0
0.30 0.35 0.40 0.45 0.50 0.55 0.60
EqLi\ederceratio

Figure 6.8 Time of flame arrival as a function of equivalence ratio.

 197

The implication from this foregoing is that, if the flammable mixture is initially rich in

the case of hydrocarbon/air then the combustion-like-explosion (bum-up) inside the

duct may be much severe than expected on consideration of the magnitude of simply

vented case, due to the high flame speeds and hence the high unbumt gas velocity ahead

of flame prior to duct entrance in the case of hydrocarbon/air. To get more insight to

this phenomena, the flow interaction between the main vessel and the duct were

illustrated in Fig. 6.9.

CH4/bj r 4) = 1.08
2-1 Erd igribon k-tcx

for Dd = 0.315 m
1.7

1.3 Pw,, ibr Dd = 0.315 m AP2-3 for Cý = 0.162

0.9
0)

0-5 Pd,, for DI = 0.162

0.1

0.1
-0.3

-0.7 Time(s)

-1.1

Figure 6.9 Flow interactions for methane/air at end ignition. (D = 1.08

The same trend observed where the maximum peak of Pma,occurred after the flame

exited the duct for both cases i. e. duct pipe diameter of 0.162 m and 0.315 m as shown

in previous chapters and there is no evidence for external explosion occurred in dump
vessel (indicated as P6 in Fig. 2.3 b). It is worth mentioning at this point that while
individual test was presented here for clarity, the results explained are indicative of the

trends displayed for this test vessel for all studied gas/air. At this point, methane/air at (D

= 1.08 would be the best representative to describe and explain the phenomena occurred
during the explosion development in comparison with the smaller duct pipe diameter
i. e. 0.162 m. The pressure difference between the explosion vessel and the vent pipe is

shown in Fig. 6.9 as a function of time, together with the vent duct explosion pressure.
 198

This shows that when the flame was in the duct, there was a negative pressure
difference,which was higher for the large vent duct attached.The pressuredifference
betweenthe explosionvesseland the duct in the initial stageof the explosioncan be
to the
used compute meanvelocity of unburntgas into the duct,during the periodbefore
the flame enteredthe vent duct. The dynamicheadpressureloss for a pipe inlet for
incompressibleflow is 0.5. The pressuredifferencereachesa maximum of 0.2 bar,just
prior to the flame entry in the vent duct, as shownin Fig. 6.9. This correspondsto a
meanunburntgasvelocity in the vent of 258 m/s. However,the sharpedgeto the vent
will produce a vena contraction and the velocity at the vena contraction, using a
contraction'coefficient of 0.61, would be 423 m/s. This is closeto the speedof sound
(360 m/s) and the compressibleform of Bernoulli's equationshould be used.However,
this shows that very high unburnt gas velocities close to sonic conditions were
generatedat the vent and thesecreatehigh turbulenceconditionsin the vent pipe which
the shock waves will be generatedand this createsa high back pressureand the
observedreverseflow back into the explosionvessel(Kasmani,Andrews, Phylaktou,
Willacy, 2007, Ponizy and Leyer, 1999a,1999b).The details of the pressureloss and
unburntgasvelocity will be discussedlater in Section6.3.

The bigger negativepressuredifferenceoccurredat the duct pipe region (illustrated as

tin and tut) indicatesthe strong outflows from the pipe to the vessel and then creates
high turbulencein the explosionvessel.Much of the unburntgasmixture remainsin the
explosion vessel at the time that the flame enters the vent duct and the turbulence
createdby the reverseflame flow from the vent duct in the explosion vesselcausesa
suddenincreasein the turbulentburning ratein the explosionvessel,marking the rate of
pressurerise to 220 bar/s and creatinga high rate of vent dischargeand thus, the large
increasein the overpressure.The higher overpressurewith the large vent was due to the
largernegativepressurebetweenthe vesseland the vent duct the large inducedreverse
flow,

The implication of this action is reverseflow causedby the larger duct diameterwould
result in largerscalemixing into the vesselas shownin Fig. 6.10, This phenomenahas
also been found by other workers (Kasmani, Andrews, Phylaktou, Willacy, 2007,
Ponizy and Leyer, 1999a,1999b).When the pressuredifferencebetweenthe vent duct
and the vesselwas high, the vent duct pressurebecamelow and this was the greatest
 199

differencefor the large vent duct. For the small vent duct the staticpressurein the duct
to
wassimilar magnitude the difference
pressure between
the and
vessel the duct.

......... .
I ........ ..
...........
1.4
......... ...........
............ ",
bu'rned gases fresh mixture

..............
...........
:

--
..............
Figure 6.10 Flow regimesin vesselbefore and after explosion-likecombustionin duct,
effect of the flame front distortion-,
a) narrow ducts b) large duct- Reproducedfrom
(PonizyandLeyer, 1999b)

that a significant increaseof

Ponizy andLeyer (Ponizyand Leyer, 1999a)demonstrated
in
overpressureoccurred relatively large vent ducts.This was shown to be due to the

compressionwave propagating towards the vessel,which was generated at the flame

front in the duct createdby the highly turbulent vent dischargeflame, A' model of the
turbulentaccelerationof dustin ductshave beenperformedby Clark and Smoot (Clark
and Smoot, 1985). The model showed that the flame acceleratesmore rapidly and
propagatesat higher velocities in larger ducts. Increasingthe diameterwill increasethe
turbulentReynoldsnumberand sincethe turbulentReynoldsnumberis the only way in
which turbulenceis incorporatedinto the model, a very high pipe turbulencelevel leads

to very high flame speedsin the vent pipe. It is consideredthat this phenomenawas
occurringin thepresentwork with the large diameterventpipe and much smallervent.

However, using the bigger duct pipe diameter seemsto agreewith the theory stated
abovefor leanmixturesconcentrationwherethe overpressureinside the vesselis lower,
comparedto the pipe havingthe samediameterasthe vent asshownin ethylene/airat (D
 200

= 0.8 for end ignition (Fig. 6.11). Apparently, the peak pressure for duct pipe of 0.162

m diameter reached its peak earlier after the flame exited the duct in comparison with

explosion with larger vent and simply vented. A possible explanation for this lies to the

different sequenceof events inside the vessel and the duct.

Q2F4/,!r 2.7
9.0
End igrition
8.0 Sn-oy
2.4
\oened
2.1
7.0 - Duct wried, Dp = 0.1 E2 m

6.0 - Duct \oertecL Dp = 0.315 m 1.8

)( Ran-e arTi\d sin-oy verted
5.0 1.5
)K Rame am\d for Dp =0 162 m
1.2
LD 4.0 x Rarne arrNed for Dp = 0.315 m Duct regocn
1 3.0 0.9
-------------------------------------- --- - ----------------
2.0 Vessedl regicn 0.6
ct
1.0 0.3

0.0 0

0.05 0.1 0-150.3

-1.0

-2.0 Tirre(s) -0.6

Figure 6.11 Pressure time histories and flame arrival for ethylene/air at (D = 0.8.

For leaner combustion at relatively small vent area, the outflow of unburnt gases into

the larger vent duct is less turbulent and hence, the combustion-] i ke-explosion event
inside the duct is less severe. Hence, the final explosion inside the vessel is reduced as

expected. Lower value Of in

Pduct larger duct diameter is due to faster discharge of fresh

from the vessel but also to earlier arrival of the flame front at the duct entrance (refer to
Fig. 6.11 and 6.13). However, it is apparent that the combustion is still occurred inside

the 0.162 m duct pipe when the peak pressure inside the vessel marked its peak,

suggesting that there is still strong interaction between the vessel and the duct before the
flame exited the duct pipe and hence, producing a greatly increased pressure rise in

respect to the simply vented explosion.

The study by lida et al (lida, Kawaguchi and Sato, 1985) mentioned that flame was
found to extinguish or hesitate in the channel before passing through in some cases,
depending on the equivalence ratio of the mixture, the channel with and the flame
inflow velocity. Other studies supported the above hypotheses by using relatively
 201

narTowductswith a sharpvessel-ductarea(Iida, KawaguchiandSato,1985,Ponizy and

Leyer, 1999a,1999b).It showedthat the flame front enteringtheduct can be temporarily
extinguisheddue to stretch and cooling through turbulent mixing with unburnedgas
which brings aboutstrongerbum-up (i.e. with higher pressureamplitudes)during re-
ignition (Ponizyand Veyssiere,2000). At this point, generallyan oscillating flow is
profoundly establishedas shownin Fig. 6.1L It seemsthat at lower equivalenceratio,
inertia of the gascolumnplays a moreconspicuouspart in pressurevariationsinside the
vesselfor the ventedexplosionwith narrowduct pipe, i.e. 0.162m diameterpipe in this
presentcase.The sameresult was also observedwith propane/airexplosionat (D = 0.8
as shown in Fig. 6.13. Even though there is no 'spiky' pressuretraces, the pressure
generatedwith duct ventedat larger duct diameteris lower comparedto simply vented
explosion,This complexsituationis difficult to explain.The possibleexplanationis due
to the reverseflow penetratesessentially into the region where there is very little
unbumedgasmixture left inside the vessel.Further, the short time for the flame front
reachesthe vesselwalls will reduceits surfacearearapidly and causeless burning rate
inside the vesselandhenceproducinglower pressure.

The currentresultsgenerallyshowedthat the use of enlargedduct diametercan reduce

the Pn,. below the value generatedwhen using the narrow duct when applied to leaner
concentration but P,,.,, profoundly higher or about the same when applied to
rich/reactiveconcentrations.From theseresults, it can be deducedthat the flame self-
accelerationis the important role in effecting the Pm'uinside the vessel with the duct
pipe attachedwhich is not be accountedin the venting guidelines i. e. NFPA 68 and
European Gas Explosion Venting Guidance 2007. Further, these results are strictly
applicablefor openventing eventhoughit is unlikely that the P,,,,,,of the vesselwill be
affectedif P, doesnot exceeded100mbar.
 -102

C2H4/ar
1.2
e Lixt vened, Dp = 0.1 62m end
1.0 igriftion
9 ElictveriKDp=0.162rTi
cet14e igrition
0.8 D-ict vertad. Dp = 0.315 ni
x
Eyid igrýtion
0.6 w Duci veried, Dp = 0.315 m
ceritre igiitici
0.4

02
.

0.0
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85
EquivaJencerclio

Figure 6.121Pressure inside the duct fOr ethylene/airas I I'LlnCtIOllOfeLlUlValClICCral'O.

O.s C'3Waj r (1)= 0.8 25

,
End ignition
0.4 Srroy %,
wted 20
DO %aled, Dp = 0.162 m
-
'- Uzt \aved, Dp = 0.315 m
0.3
)< Flame am\d sin-Oy \.eled ý 1.5
)K Ramearri\d, Dp=0.162m Dud region

)( FIamearri\d, Dp=0.315m
0.2 1.0
-----------------------------------
Vessd region
0.1 0.5

0 ý,- --flir 0.0

0.05 0.1 0.15 0.2 0.25 0.3

-0.1 -0.5
Time(s)

FigUrc6.11 Pressuretime historiesand flanic arrival for propane/airat (1)= 0.8 for end
ignition.
 203

6.2 Flame speedsanalysis

Figure 6.14 showsthe averageflame speedsmeasuredinside the vesselandthe duct for

methane/air.As reported from previous work (Kasmani, Andrews, Phylaktou and
Willacy, 2007a,Kasmani,Andrews,Phylaktouand Willacy, 2007b), the induced flow
throughthe duct plays an importantrole in the final severityof the explosion.This flow
in the vent pipe is driven by the flame speedsand burnt gas expansionin the main
vessel.The samefigure also showsthat at lean methane/airmixtures, i.e. (D = 0.68,
there was negligible effect of the duct length or diameter (L/D ratio) on both flame
speedsinside the vesseland the duct. At (D = 1.08,the highestexplosion vesselflame
speedof 22.8 m/s occurredfor the short vent duct, which was closeto a free discharge.
For the largevent duct (Dp= 0.315m or LID = 3.17)the peakupstreamflame speedwas
19 m/s andfor the vent duct the samesizeasthe vent (L/D = 6.17), it was 16.8m/s. The
flame speedsapproachingthe vent were much lower for lean mixtures and hencethe
inducedflow waslower andthe impacton the overpressures
of the vent ducts were then
lower.

On
--4- DuctVented,
mendignihon
Dp 0.315
CH4/aIr b)
Duelvented, I)p 0.316nk 600- CH41air
CentreIgnition -4- Dp - 0.315m, end
Ignition
-X- Ductvertled,Do. 0.162m -ý 500 Do - 0.315m, tentral
andignton I IRoo
milon,
-*-Duct vernedI)p- 0.162n a - 0.162 m, ond
CentreIllnition 400 - -a-
-0 Ignition
SwO vented, end Onown
-*- Dp - 0,162,conlrid
300 ignition
ý50" Vented,Caml,m)
200

100

0.4 0.5 0.6 0,7 0.8 0.9 1 1.1 1.2 0.4 0.5 0.6 0.7 0,11 0.9 1 111 1.2
Soulvalanceratio Equivalenceratio

Figure 6.14 Averageflame speedsmeasured(a) in secondhalf of the vessel(betweenTj

and T3 as in Fig.2.3b) and (b) in the vent duct (betweenT4 and Tg) as a function of
equivalenceratio for methane/air.

The flame speedsin the main vesselapproachingthe vent were considerablyhigher than
for sphericallarninar flame speeds.This was due to two effects:firstly, self acceleration

of the flame fluough the cellular flame front mechanismand secondly,the suctioneffect
 204

of the vent dischargeon the flame shapewhich would draw the flame expansion
preferentiallyin the direction of the vent (Kasmani,Andrews,Phylaktou and Willacy,
2007a, Kasmani,Andrews, Phylaktouand Willacy, 2007b). These effects were both
higher for end ignition as the distanceto the vent was double that for central ignition
and hence the two effects were enhanced.This was the main reason why the end
ignition gavethehighestoverpressures
comparedwith centralignition.

The flame speedsinside the duct were similar for both duct diameters,apart from the
equivalenceratio at (D= 1.08wherethe highestflame speedsof 490 m/s experiencedat
LJD = 3.18.The larger vent duct createda flow expansionfrom the vena contractionat
the inlet vent to the duct wall. This flow expansioncreatesa pressureloss that is the
sourceof the turbulencethat accelerates the flame which is largerwhen there is a larger
flow expansion.For the presentgeometry,the unbumt gas dynamic head pressureloss
in termsof the vent areadynamichead,was0.47 for the 0.162m vent pipe and 1.27 for
the 0.315 m diametervent pipe. This producedmore turbulenceand a greaterflame
accelerationof the flame inside the larger vent pipe, as shown in Fig. 6.14. Also the
lower meanvelocitiesin the larger pipe would enablea flame to propagatein regions
wheretherewaslocal turbulent quenchingin the smallervent pipe (Kasmani,Andrews,
Phylaktouand Willacy, 2007a,Kasmani,Andrews,Phylaktouand WillacY, 2007b) and
this would increasethe back pressure,as found experimentally.For leaner mixtures the
velocities were much lower and the turbulencegenerationwas significantly lower as
this is proportionalto the squareof velocity. Hence, the effect of the vent pipe was
muchlower for the slowerburning leanermixtures.

However, the trend for leaner concentrationresults in lower flame speedsin the duct
seemsnot applicablein the caseof hydrogen/airexplosion.As illustrated in Fig. 6.15,
the flame speedsinside the duct was higher for all equivalenceratio for vent duct the
samesize as the vent comparedto 0.315m diameterduct for end ignition and marking
its highestvalue at 484 m/s. This high flame speedsare due to the comparatively fast
flame from the vesseldue to large flame areaeffect. Studyby Aung et al (Aung, Hassan
and Faeth, 1997)showedthat hydrogenat lower equivalenceratio 0.3 < (D < 0.6 are in
unstablepreferential-diffusionconditions where developedchaotically irregular name
surfaces.
 205

a) (b)
H21air
1 -4- " veriled.endgnam fiýair 600
100 Ductvented.
Dp- 0,162m,
end
Z
90
-"
Ignition
w4d'sawpow
500
-4-DUCIVOnted.00-0.142m,
'm I
voted, DP-0.162lkWd cenkoIgma"
-*-Dxl
70 4"Alo J? D-uclvented,on 0,31Sn end
400
Dict, ol-G. Dp- 0,162mcatwe ig*on
60 :p
Ductvented,
Dp 0.315m,
50 300
contisignAmm
40 1 ýDXIVWW np-0315M emit
200
a 30

20 100
r
w 10
0
0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Equivalence
ralio EquWlertce
ratio

Figure 6.15 Flame speedsmeasureda) in secondhalf of the vessel(betweenTj and T3

as in Fig.2.3b) andb) inside the duct (betweenT4 and Tg) as a function of equivalence
ratio for hydrogen/air

The resultsagreedthat Mark-steinno, Ma = -1 at (D= 0.3 with neutral conditions(Ma =

0) at 0=0.7. Markstein number is a physiochemicalparameterthat expressesthe
responseof a flame to stretching.At low Ma no, all sphericallyexpandingflames are
intrinsically unstable with no stabilizing influence due to thermodiffusive effect
(Bradleyand Harper, 1994).As the flame is stretched,it beginsto weakenand recede
downstreamdue to the reducedlocal burning speed.This patterncontinuesto develop
in time until the flame sheet finally breaks(Bell, Cheng, Day and Shepherd,2007).
Flame stretch influenceslaminar burning velocities through preferential diffusion of
heat and mass.Within the thin-flamelessregime, this can causethe instability of the
flame surface dependingupon the sign of the Markstein number. If the Markstein
numberis negative,the laminar burning velocity increasesasthe flame stretchincreases
with the following equation,

SL-ISL =1+ MaKa (6.2)

where SL_is the value of larninarburning velocity when the flame stretch,K=0, SL is
the Ian-tinarburning velocity, Ka is the Karlovitz number and Ma is the Markstein
numbergiven by equationsasbelow respectively,

Ka = K,61ý/ SL (6.3)

8D=
D. 'SL (6.4)
 206

Ma = L1,5D (6.5)

statedin the equationsarerepresented

The parameters as;

K= flame stretch

SD= characteristicof flame thickness

Du = massdiffusivity

L= Marksteinlength

DimensionlessKarlovitz and Markstein numbers are defined to characteriseflame

stretchand the responseof a flame to stretch.For unstablecondition, finite levels of
stretchcauseflame temperaturesto be higher than for the unstretchedflame (roughly
1550 K for the stretchedflame with respectto 1400K for unstretchedflame) (Aung,
Hassanand Faeth, 1997). Perturbationof lean, premixed hydrogen flame induces a
growing instability that evolvestowards a cellular structurein correspondingto the
increaseof flame temperature,forming cracks or cuspsin the flame surfaceand this
cross-crackingdevelopsuntil eventuallythe flame surfacebecomescompletelycellular.
This condition then promotesfaster reaction rates and thus higher laminar burning
velocitiesfor the stretchedflame. This is associatedwith a marked increasein flame
speedsandthis condition agreedwith the experimentresultsshowed.The schlierencine
photographin Fig. 6.16 shows the cellular flame developmentduring explosion of
H2/airmixturesat (D= 0.26.

Even thoughP,,,,,for larger duct pipe at centreignition exhibited highestvalue (refer to

Fig.6.7), but yet the flame speedsinside the vesselandthe duct disagreedwith the result
'.
asP.,, crS,, Theseresultsapparentlyshowedthat for centreignition, the link between
the vesselandduct pipe seemsto be independentat somepoint, wherethe final value of
P.. dependson the substantialamount of unbumt gasesleft inside the vesselas been
postulatedby others(Ferrara,Willacy, Phylaktou,Andrews,Benedetto,Salzano,Russo,
2008,PonizyandLeyer, 1999a,1999b).
 207

A similar effect would occur for smaller K or larger vent diameters.The vent flow
velocitieswould be reduced,lower turbulencelevels in the pipe would be generatedand
lower overpressure
increasesdue to the useof a vent would occur. It is significant that
much of the publishedwork on vent pipes, which all showsa large increasein the
overpressurecomparedwith a free vent discharge,was all carried out with relatively
large valuesof K. More works on vent pipes has beencarriedout for dust explosions,
but generallywith K> 10. It is consideredthat in view of the limited experimentaldata
on the impactof a ventduct on the overpressurefor gaseousexplosionand the potential
importanceof K and mixture reactivity (which determinesthe vent flow velocity) that
more works is requiredto understandthis type of venting phenomenaand to provide
morereliableventingdesignguidance.

6.3 Pressurelossand unburnt gasvelocity

As mentioned earlier in Section 6.1, the pressure loss prior to the duct pipe entrance
(indicated as AP2-3)and inside the pipe (AP3-6),will be used to calculate the unbumt gas

velocity prior to duct entrance as well as the gas velocity inside the duct to ascertain the
enhanced burning rate due to turbulisation and enhanced pressure drops due to the
geometry are playing the important role to the final pressure inside the vessel. This can
then leads to the comparison of the pressure loss for the vent pipe as being the same as
the vent orifice and larger vent pipe. This simple analysis can be evaluated using;

APi,,+ APd,,,,
+ AP.,,,=1 pSg'(Kin + 4f -ýL+ Ni + K,,,,,) (6.6)
2D
 I ()1ý1'

tit ;I
ýI.,.

II

(III
. .1ýý
,I--I,
11-I'laMCdCVCIOjIIIICllt
Figurc 0.16 Sclifici-cii cinc photographS01'CCIIUI, CXIIIOSiOll
(IL11-ilig

of' H-,/air IIIIXILII'C,(1) = 0.26. Reproduced 1'rom Bradley and Harper (Bradley and
Harper, 1994).
 209

Ki,, and K t are respectively the pressureloss coefficients for sudden flow area
..
(in
restriction/enlargement the caseof a flow to larger pipe from the vessel)and sudden
flow areaenlargement(pipe to dump vessel).f is the friction factor for he flow inside
the duct evaluatedfrom Darcy-Weisbachequation,Ni is the velocity heads lost in
fittings in this presentcasefor duct pipe of 0.162m,one gatevalve and two couplings
in the duct; valuesreportedin (Ferrara,Willacy, Phylaktou, Andrews, Benedettoand
Salzano,2005)wasused.Below is the list of parametervaluesused;
Kin = 0.5; K t=0.95; f=0.005; 1: N; = 0.25
...

The pressuredrop at the entranceof the duct pipe canbe expressedas,

1
APin =K psg (6.7)
2

The actual value of K is strongly dependenton the geometry of the component

considered.It may alsobe dependenton the fluid properties.That can be said,

O(geometry,
Re) (6.8)

where Re is the pipe Reynoldsnumber given by Re = pSRD /,a . For many practical

applications,the Re no is large enoughso for the flow inside the pipe where inertia
effect is more dominated rather than the viscous effects, is usually found that the
pressuredrops and head lossescorrelatedirectly with the dynamic pressure.Thus in
most cases,the loss coefficientsfor componentsare a function of geometryonly as Eq.
6.8 will be K= 0(geometry)(Munson, Young and Okiishi, 2006). Meanwhile, for
lossinsidethe pipe,
presSUTe

I
2 L+
PSv (4f Nj) (6.9)
2D

From Eq.6.6,6.7 and 6.9, unburnt gas velocity on each phasecan be evaluatedand
determinedaslisted in Table6.2.
 210

Table6.2 Calculatedunburntgasvelocity at the duct entrance,inside the duct pipe and

at the duct pipe exit for studiedfuel/air mixtures.
FueL/air Pipe Ignition q) S, at duct S,,inside S. exit,
diameter, position entrance the duct (M/S)
Dp (m) (M/S) (m/S)
CH4/air 0.315 End 0.8 158.9 415.6 145.4
1.05 251.0 942.3 357.5
Centre 0.8 60.8 236.0 83.1
1.05 95.4 769.6 222.2
0.162 End 0.8 129.3 573.9 305.5
1.05 235.8 634.2 367.5
Centre 0.8 59.0 399.3 167.7
1.05 103.6 501.5 2?2.7
C3Hg/air 0.315 End 0.8 70.2 225.5 94.3
1.0 305.2 823.5 320.3
1.375 260.4 1437.5 505.95
Centre 0.8 55.1 169.5 47.8
1.0 166.7 548.7 177.6
1.375 113.6 1379.7 333.1
0.162 End 0.8 138.4 425.8 245.7
1.0 254.1 686.7 429.4
1.375 243.8 671.3 429.1
Centre 0.8 69.0 206.9 113.9
1.0 109.3 559.0 365.4
1.375 109.3 443.4 296.4
C2114/air 0.315 End 0.8 260.1 879.1 331.3
Centre 0.8 111.4 1574.5 200.5
0.162 End 0.8 207.7 562.1 508.1
Centre 0.8 150.1 436.2 372.9
112/air 0.315 End 0.54 246.9 1470.2 313.6
Centre 0.54 145.9 1820.5 391.9
0.162 End 0.54 268.4 688.9 554.3
Centre 0.54---ý 132.2 434.0 252.1
 211

From the tabulateddataabove( table 6.2), someinterestingconclusionscan be drawn.

The high S9at the duct entrancecausesthe restrictedflow to the duct pipe and this
eventuallyleadingto the chockedflow (sonicventing)due to the higher turbulentfield
in the initial of the duct entrance.This high jet flame velocities produce remarkably
high velocitiesof unburnedgasesinside the duct where 0.315 m diameterduct pipe,
markingto attain highestS. inside the pipe for rich concentrationfor methane/airand
propanelairwith respectwith 0.162 m diameterduct pipe at the same concentrations.
Interestingly,the very high S, inside the pipe observedfor H2/air at (D = 0.54 for both
ignitionsin the caselargerpipe diameterwhich is Sg> 1400m/s.This high unburnt gas
velocity reflectsto the high flame speedsif the relationshipof Sf = Sg+ S,,is used. At
0=0.54, S,, is 1.0 m/s (Andrews and Bradley, 1973) and from the mentioned
relationship,Sf inside the duct is 1471 m/s and 1821 m/s for end and centre ignition
respectively.This high flame speedsinside the duct is very closeto the CJ-velocity of
1970ni/s (Dorofeev, Bezmelnitsinand Sidorov, 1995) which the onset of detonation
condition.From this, turbulenceinducing elementscan leadto a significant acceleration
of the flame front inside the vessel due to the backflow and flame instabilities and
consequently,to more intense mixing of unburnedmixtures still present inside the
vesseland thus, rapid rise in final pressurein the vessel.Due to the abovementioned
condition,it is clear that venting at larger duct pipe diameterat lean concentrationi. e.
16 % for hydrogen/airandrich concentrationsfor low burning velocity mixturesi. e. 5.5
% propane/airare less effective in reducing the peak pressureinside the vessel in

comparisonwith the narrowduct pipe attached.

From this analysis,it canbe confirined that the high unburritgasvelocity inside the duct
induceda very high turbulent level and thus a severesecondaryexplosion inside the
duct.The secondaryexplosionin the duct which in turnsaffectsthe residualcombustion
in the main vesseland showsmore violent for endignition as shownin Fig. 6.4 and 6.5.
However,the higher intensity of the secondaryexplosion does not affect literally in
practisethe final pressureinside the vesselin the caseof central ignition especiallyfor
high burning velocities mixtures i. e. hydrogen/airand ethylene/air.It is postulatedby
others (Ferrara, Willacy, Phylaktou, Andrews, Benedetto, Mkpadi, 2005, Ferrara,
Willacy, Phylaktou, Andrews, Benedetto,Salzano,Russo, 2008, Ponizy and Leyer,
 212

1999b)that chemicalcontributioni. e. mixture reactivity left inside the vesselplays an

importantrole to the pressurerise insidethe test vesselasshownin Fig. 6.6 and6.7.

6.4 Concludingremarks

Enlarged vent ducts, i.e. ducts with areas greater than the vent area gave vent
similar to those for free vents for lean mixtures,but for the maximum
overpressures
reactivity mixture ((D =1.08 for methane/airand (D = 1.375for propane/air)P ..a, was
similar to that with a vent ductthe samediameterasthe vent.

The causeof the large increasein overpressurefor both ducts in the case of rich
concentrationfor methaneand propane/airmixtures was due to the high induced
unburnt gas velocity into the vent and inside the vent duct itself. For the presentK=
16.4,this conditioncreatednearsonicflow conditionsat the vent venacontraction.The
arrival of the flame in the vent createdsonic flow in the vent duct and the high back
pressurecreateda reverse flow into the explosion vessel.This high turbulent flow
acceleratesthecombustionof remainingunburnt,mixture andthis further acceleratesthe
flow in the ductcreatingthe peakoverpressure.

It can be saidthat the ignition position plays an importantrole in determiningthe final

pressureinsidethe vesselin which end ignition gave highestP,,,. for methane/airand
propane/airin all cases.However,for high burning velocity mixtures i. e. hydrogen/air,
the trend is seemsnot to have a good agreementwhere central ignition gave highest
P.. with a profoundly 'spike detonation'peak. The major finding is the occurrenceof
auto-ignition at the unburned pocket of mixtures trapped inside the vessel and it
contributes a significant role in the hydrogen/air explosion developmentbut not in
methane/air.It then can be postulatedthat with central ignition, the bigger quantity of
fresh mixturesleft for the residualcombustionin the vesselis the main contribution of
the high magnitudeof 'spike detonation'pressurein comparisonwith the pressurepeak
attainedat endignition inside the vessel.

The presentdesigncorrelationsfor explosion ventsand vent ducts,which are basedon

very limited experimentaldata,do not predict the presentresultsand their reliability for
small vesselvolumeswith high K is in doubt. Furtherwork is requiredin the areaof the
 213

impact of K, mixture reactivity and static burst pressureon vent design with vent
dischargeductsattachedif morereliabledesignguidanceis to be given.
 CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

FOR FUTURE WORKS
 214

7.0 Summary of major rindings

General effects of venting explosions

The initial project was to compileand collect all publishedexperimentaldata available

for different gasreactivity in ventingexplosionin order to investigatethe validity of the

publishedventingcorrelationsin relationto the venting practicality. Although abundant

data and several excellent reviews are available, there are no universally agreed
correlations for guidance on the selection of vent areas. The most used and
recommendedcorrelation for venting gas deflagration is adopted by NFPA 68
(NFPA68,2002) for low and high-strengthenclosures basedon the works of Swift-
Epstein(Swift, 1983) and Bartknecht(Bartknecht,1993) respectively.It is found that
the correlationderivedby Bartknecht(Bartknecht,1993)have beenshownto be under
predictedmost of the datapresented.It can be said that the failure of the Bartknecht's
correlationis due to the assumptionof the samevent areais requiredirrespectiveof the
volume. Further, most of the vesselsinvolved in Bartknecht's correlation were large
wherethere is suspectedthat self-accelerationmay occur during the ventedexplosion
and this parameterdid not be takeninto accountin an appropriatemanner.However, it
is recommended that the validity andlimitation of Swift's equation(Swift, 1983)can be
extendedto wider rangefor P,,d > 200 mbarproviding the parameterP, is addedinto the
equation.

The massburning rate of the sphericalflame propagationequationand incompressible

flow equationwereusedin orderto correlatethe P,,d and A, for venting designpurpose.
From this theory, two methodswere derived namely as Method I and Method 2. The
equationgiven respectivelyas,

m, = AS,, p,, = CdCA,(2p,,P,.,

d)o"
(7.1)

m = ASp = ASý,(E - 1)p = CdeA,(2pPd)0-5 (7.2)

It can be recommendedthat Method 2 gives reasonablygood agreementwith most of

the experimentaldata and the use of S, term to describethe unburnt gas displacedby
the flame which gives approximately-6 timesthe massflow rate suggestsit gives close
estimationon P,.,d in relation with practical application in comparisonby with S,,term
 215

usedin Method 1. The net effect is as Sgis close to the flame speed,Sf in value, the
approachis only slightly lower ventmassflow ratethanthat basedon Sf.

A significant flame self accelerationeffect for subsonicventing was shown for K<-5
and this effect is similar to vent inducedturbulenceand could be accountedby the P
term in the burning velocity equation,The turbulenceenhancementpredictedbasedon
Bartknecht'sequationand proposedMethod 2 was in a good agreementwith P derived
from tabulatedexperimentaldata. It can be said that the 0 derived was perfectly
reasonablevaluefor P usedby other works (Munday, 1963,Pasman,Groothuizenand
Gooijer, 1974,Yao, 1974).However,the flame experiencesdecelerationeffect in larger
volumeswhenpressureis high in larger K and this effect has never beenhighlighted
previously. It is postulated that at high K and Predwith sonic venting during the
explosions, the self-accelerationis likely to have already occurred at the smaller
volumes.

It can be suggestedthat the use of K term is more suitableto be applied on cubic and

sphericalvesselfor LID =2 but failed to give satisfactoryresultsfor non-cubicvessels.

The AdA, term is more favourableto correlatethe influenceof vessel'sgeometriesfor
non-cubic vessels.The data and figures shown in this work also illustrated that Pmd
exponent of 0.5 in the subsonicflow regime and unity in sonic flow regime are
supportedby the experimentaldata. The main data scatter is due to the additional
influence of volume, The use Of Predexponent of 0.582 in Bartknecht's equation
compromisesbetweensubsonicand sonic flow regime which gave most of the data
scatteredoutsidethe line.

7.1.1 Comparison of theory and experimental results

Experimentsin two different cylindrical vesselvolumesin this presentstudyhave been

usedto identify the physicsanddynamicsmechanismsresponsiblefor the generationof
the pressurepeaks,in particular on the occurrenceof the deflagration-to-detonationin
simply ventedexplosions.Generally,end ignition giveshigher P,,. comparedto central
ignition in hydrocarbon/airmixtures.For end ignition, the name is allowed to propagate
in one direction,leading to more elongatedflame towardsthe vent and hence,increase
in massburningrate and high flame speed.Self-accelerationis expectedto be one of the
 216

importantfeaturesfor the increasein P.,, magnitudewhich occurredin Test vesselI in

comparisonwith Pn,. in Test vessel2 at the sameequivalenceratio. This is justified by
the reversedcalculationdone usingMethod2 equation.It shownthat about80-90 % of
flame area has been occupying the vessel total surface area. It is confirmed the
observationreportedby McCann et al (McCann, Thomas and Edwards, 1985). In their
work, they saidthat the flame cellularity (self-acceleration)appearedin earlier stagein
largervolumeand this give significant effect on the overpressure inside the vessel.For
i.
smaller vessel e. 0,0065 3,
m. experimentshave shown that the presenceof pressure
oscillation,coupling with the inducedturbulenceby the vent flow, increasedthe Pn,a)'.
For centrally ignited explosion,the increasedintensity of the flame cellularity during
flame propagationproducesacceleratingflame front which later interactwith the vessel
wall. Due to the rapid decelerationof the flame front as it approachthe vesselwall, it
resultson the strongrare fraction waveswhich triggering further combustionof a large
amountof unburntgases left inside the vesseland lead to a significant P.,,,,inside the
vesselfor reactivegas/airmixturesas shownon ethylene/airandhydrogen/airmixtures.

Auto-ignition is the main factor of the appearanceof spiky pressure traces on

hydrogen/airandethylenelairin the test vessels.It can be saidthat fast turbulentmixing

of the combustionproductsand reactantsinitiates the 'hot spot' or auto-ignition centre

leadingto the transitionto detonationexplosion.

7.1.2 Duct vented gas explosion

The high overpressuredue to the addition of a vent duct, in a 0.2 m3 cylindrical vessel
with an LID of 2 and K of 16.4,occursafter the flame has exited the vent duct, but is
not due to an external explosion. There is found that substantialamount of unburnt
gasesJeft insidethe vesselafter the vent burstsis the leadingfactor in increaseof P. ý,"
for high burning velocity mixtures at centrally ignited, The associategas velocities
aheadof the flame createhigh unburntgasflows conditionsat entry to the vent and this
give rise to high back pressures.

As P, increases,the distanceof normal sphericalflame propagationincreasesand there

is a further reductionin accelerationdistance.This initially reducesthe overpressureat
low P, The effect of the vent burst pressureis to increasethe flow velocity in the duct
 217

when the vent burst,as the flame has had more time to grow upstreamof the vent and
this givesa highervent duct flow velocity oncethe vent cover bursts,The effect of the
vent burst pressureis complex and non-linear and was not representedby the linear
effect in the correlationof Bartknecht(Bartknecht,1993)usedin NFPA 68 (NFPA68,
2002).

The greaterself accelerationof flames due to cellular flame developmentfor propane

relative to methaneis shown in the results. Propanehas a very strong influence of
equivalenceratio on the flame speedsand overpressures,especiallyfor rich mixtures
wherethe highestflame speedsand overpressuresare exhibitedfor (D = 1.35.It cannot
be assumedthat the mixture with the highest laminar burning velocity, measuredon
small diameterflames with no cellular flames, is the mixture with the worst case
explosion hazards.For gases such as propane that have a strong cellular flame
developmentthe worst caseexplosionhazardwill be for rich mixtures,wherethis effect
is maximized.In the presentwork this was (D = 1.35 for propane.For methanethe
vented overpressureoccurredat the same mixture strengthas that for the maximum
larninar burning velocity, but the cellular flames did contribute to the higher flame
speedsand overpressures. This is becausefor methanerich mixtures do not have
greatercellularity.At lean concentration(low equivalenceratio), high burning velocities
mixturesexhibit a detonationspike in the pressuretracesinsidethe vesseleventhough
the behaviourof secondaryexplosion( burn-up) andpressuredrops at the duct entrance
reproducewhat havebeen observedin methane/airand propane/airmixtureswell after
the flame hasleft the vessel-ductassembly.

Enlarged vent ducts, i. e. ducts with areas greater than the vent area gave vent
similar to those for free vents for lean mixtures, but for the maximum
overpressures
reactivity mixture ((D =1.08 for methane/airand 0=1.375 for propane/air)P,,,,,was
),
similar to that with a vent duct the samediameterasthe vent.

The causeof the large increasein overpressurefor both ducts for rich concentrationin
low burningvelocity mixtures wasdue to the high inducedunburntgasvelocity into the
vent and inside the vent duct. For K= 16.4, this condition created near sonic flow
conditionsat the vent venacontraction.The arrival of the flame in the vent createdsonic
flow in the vent duct and the high back pressurecreated a reverse flow into the
 218

explosion vessel.This high turbulent flow acceleratesthe combustionof remaining

unburnt mixture and this further acceleratesthe flow in the duct creating the peak
overpressure.

It can be said that the ignition position play an important role in determiningthe final
pressureinside the vesselwhereend ignition gave highestPn,,,,with respectto central
ignition for methane/airandpropane/airin all cases.End ignition gives greaterdistance
towardsthe vent andhence,theflameacceleration overtwice thedistanceasfor
continuous
the centralignition,However,for high burning velocity mixturesi. e. hydrogen/air,the
trend is seemsnot to havea goodagreementwith centralignition gavehighestPmaxwith
a profoundly 'spike detonation' peak. The major finding is the occurrenceof auto-
ignition at the unburnedpocket of mixturestrappedinside the vesselwhich eventually
playing a significantrole in the hydrogen/airexplosionsbut not in methane/air.It then
can be postulatedthat at central ignition, the bigger quantity of fresh mixtures left
inside the vesselmarked as the main contribution of the high magnitudeof 'spike
detonation'pressurein comparisonwith the pressurepeakattainedat endignition inside
the vessel.

7.2 Recommendationfor the future work

From the discussionon the importantparametersaffecting the pressuredevelopmentin

vented gas explosion, some conclusionshave been highlighted. The practicality of
Bartknecht'scorrelationis only in good agreementwith experimentalresultsfor K<- 5
as shownin Chapter3 andthis equationhasbeenadoptedby NFPA 68 (NFPA68,2002)
and EuropeanStandard(2007). From simply vented experiment,the Bartknecbt's and
Swift's equationsgave gross estimation in comparisonwith tile experimentalresults
obtainedwith both 0.2 and 0.0065 m3 cylindrical vessel.In the discussion,it is also
postulatedthat self-accelerationplays a major role in an increaseof P,,d in vented gas
explosionfor largevolume.It is also found that vessel'sshape/geometry influencing the
final Predinsidethe vessel.In orderto validatethis indicateddependenceandto improve
the prediction of current vented gas correlations, experimentsinvolving lower K in
cylindrical andsphericalvesselsat bigger scalecould be carried out in the presenttest
facility i. e. V=Im3 in sphericalvesseland 5 M3 in cylindrical
vessel.This is also to
 219

validate and quantify the A,/A, term to P,,,,, in venting design insteadof K for non-
cubic vessel.

The presentdesign correlationsfor explosion vents and with vent ducts, which are
basedon very limited experimentaldata, do not predict the presentresults and their
reliability for small vesselvolumes with high K is in doubt. In order to justify the
reliability and applicabilityof the currentdesigncorrelationon this subjectmatter,more
work on smaller vesselwith high K with different mixture reactivity are needed.The
useOf instead
Pdynamic of Pvis morefavourable in orderto correlatethe influenceof vent
cover to Pmax, dP/dt and flame speeds.It is recommendedto use commercial vent
coverfor future work in orderfor vent coverinertia effect to be taken into account.

The experimentalresults gave variation of mixture reactivity and it observed that

hydrogen/airand ethylene/air(high burning velocity mixtures) behaveddifferently with
propane/airand methane/air(low burning velocity mixtures). At certain condition,
detonation spike is detectedin the tests for hydrogen/air and ethylene/air. It is
interestingto study other high burning velocity mixtures such asbutaneand penteneas
they are both has higher laminar burning velocity than propane/airand propensity for
flame cell cellularity is likely to occur at lower concentrations.Further, both mixtures
aregiven little attentionon its behaviourandmechanismin venting gasexplosion.

Cameraand Schlierenrecordingscould be fitted inside the test facility in order to get

more precise information regarding the mechanism developed during vented gas
explosion,Further work is requiredin the areaof the impact of K, mixture reactivity
and static burst pressureon vent design with vent dischargeducts attachedif more
reliabledesignguidanceis to be given.

Further, the essentialrole of substantialamount of unburnt gas left trappedinside the

vesselthat hasbeendiscussedin detail could be benefitedfrom the applicationof CFD
(Computational Fluid Dynamics) modelling. Previous CFD modelling (Ferrara,
Benedetto,Salzanoand Russo,2006) showedthat the main mechanismaffecting the
pressurerise during gas explosionin duct-ventedvesselis the violent combustionthat
occurs in the initial section of the pipe which also been one of the major factors
discussedin this project. However, more quantitative predictions for a more realistic
 220

modelof gasflow needto be introducedin orderto take into accountthe fluid dynamic
effect and the dynamicsof flame propagationand associatedpressureand temperature
during the ventedgasexplosion.Further,to mitigatethe severityof unbumt gasvelocity

at the duct, the useof flame arresterat the venacontractaregion would be the solution
to minimise the intensificationof secondaryexplosioninside the vesseland hence,less
physicalbackflow to the decreasing
vessel, the final inside the vessel.
overpressure
REFERENCES
 221

RERERENCES

EuropeanStandard:GasexplosionventingguidanceEN 14994:2007.
Abdel Gayed,R. G. & Bradley, D. (1981) A two-eddy theory of premixed turbulent
flamepropagation.TransProceedingsof RoyalSocietyof London,A 301,1-25.
Andrews, G. E. (2004) Course notes on gas and dust explosionprotection design,
Universityof Leeds,United Kingdom.
Andrews, G. E. & Bradley, D. (1972) Determinationof burning velocity: A critical
review. Combustionand Flame, 18,133.
Andrews, G. E. & Bradley, D. (1973) Determinationof burning velocity by double
ignition in a closedvessel.Combustionand Flame,20,77.
Andrews, G. E., Bradley, D. & Lwakabama,S. B. (1975) Turbulenceand turbulent
flamepropagation-acritical appraisal.Combustionand Flame,24,285-304.
Anthony, E. J. (1978) The use of venting formula in the design and protection of
building and industrial plant from damageby gas or vapor explosions. Journal of
HazardousMaterials, 2,23-49.
Aung, K. T., Hassan,M. 1. & Faeth,G. M. (1997)Flame stretchinteractionsof laminar
prernixedhydrogen/airflames at normal temperatureand pressure,Combustionand
Flame, 109,1-24.
Bartknecht,W. (1985) Effectivenessof explosionventing as a protective measurefor

silos.PlantlOperationalprogress,4,4-12.
Bartknecht,W. (1993)Explosions-schultz, Berlin, Springer-Verlag.
Bartknecth, W. (1981) Explosions course prevention protection, Berlin,New York,
Springer-Verlag.
Bell, J. B., Cheng,R. K., Day, M, S. & Shepherd,I. G. (2007) Numerical simulation of
lewis numbereffectson leanpremixedturbulentflames.Proceedingsof the Combustion
Institute,31,1309-1317.
Bradley, D. (1999) Self accelerationof flames. 2nd Fire and F-TplosionsHazards
Seminar.
Bradley,D., Cresswell,T, M. & Puttock,J. S. (2001)Flame accelerationdue to flame-
inducedinstabilitiesin large scaleexplosions.Combustionand Flame, 124,551-559.
Bradley, D., Emerson,D. R. & Gu, X. J. (2003) Modes of reaction front propagation
from hot spots.Combustionand Flame, 133,63-74.
 222

Bradley, D, & Harper, C. M. (1994) The developmentof instabilities in Jaminar

explosionflames.Combustionand Flame,99,562-572.
Bradley,D., Hicks, R. A., Haq, M. Z., Lawes,M., Sheppard,C. G. W. & Woolley, R.
(2003)Turbulentburning velocity, burnedgasdistributionand associatedflame surface
definition. CombustionandFlame, 133,415-430.
Bradley,D., Lau, A. K. C. & Lawes,M. (1992)Flame stretchrate as a determinantof
turbulentburningvelocity. Phil.Trans.R.Soc.Lond.(A), 338,359-387.
Bradley, D. & Mitcheson,A. (1978a) The venting of gaseousexplosionsin spherical
vessel(1)Theory,Combustionand Flame,32,221-236.
Bradley, D. & Mitcheson,A. (1978b) The venting of gaseousexplosionsin spherical
vessels(11)-theory
andexperiment.Combustionand Flame,32,237-255.
Bradley,D., Sheppard,C.G.W., Woolley, R., Greenhalgh,D.A. & Lockett, R.D. (2000)
The developmentand structureof flame instabilities and cellularity at low Markstein
numbersin explosion.CombustionandFlame,122,195-209.
Buckland, I. G. (1980) Explosions of gas layers in a room size chamber. 7th
InternationalSymposiumin ChemicalProcessHazardswith special referenceto plant
design.Manchester,I.Chem.E.SymposiumSeriesNo.58.
Burgoyne,J. H. & Wilson, M. J. G. (1960) The relief of pentenevapor-air explosion in
vessels.L ChemE.Symposiumon ChemicalProcessHazards,25.
BS 1042,Part 1: Section 1.5: Measurementof fluid flow in closed conduits. pressure
differentialdevices( 1997)
Canu,P., Rota;R. & Carra,S. (1990) Ventedgasdeflagrationsa detailed mathematical

modeltunedon a large setof experimentaldata.Combustionand Flame, 80,49-64.

Chippett,S. (1984)Modelling of venteddeflagrations.Combustionand Flame, 55,127-
140,
Chow, S. K., Cleaver,R. P., Fairweather,M. & Walker, D. G. (2000) An experimental
study of ventedexplosionsin a 3: 1 aspectratio cylindrical vessel.Trans IChemE,Vol
78, PartB, 425-433.
Clark, D. P. & Smoot,L. D. (1985)Model of acceleratingcoal dust flames.Combustion
and Flan2e,62,255-269.
Clarke, A. (2002) Calculation and considerationof the lewis number for explosion
studies.Trans.IChemE,80, PartB, 135-140.
Cooper,M. G., Fairweather,M. & Tite, J. P. (1986) On the mechanismof pressure
generationin ventedexplosions.Combustionand Flame,65,1-14.
 223

Cousins,E. W. & Cottons,P. E. (1951) Design closed vesselsto withstand internal

explosions,ChemicalEngineering,58,133. ,
Cubbage,P. A. & Marshall,M. R. (1972)Pressuresgeneratedin combustionchambers
by the ignition of air-gasmixtures.I. Chem.E SymposiumSeriesNo.33,33,24-3 1.
Cubbage,P. A. & Marshall,M. R. (1974)Explosionrelief protectionfor industrial plant
of intermediatestrength.L Chem.E.SymposiumSeriesNo.39a,
Cubbage,P. A. & Simmonds,W. A. (1955) An investigationof explosionreliefs for
industrial drying ovens.I-top reliefs in box ovens,London,The Gas Council: Research
CommunicationGC23.
Cubbage,P. A. & Simmonds,W. A. (1957) An investigationof explosion reliefs for
industrial drying ovens, II-back reliefs in box ovens: Reliefs in conveyor ovens,
London,The GasCouncil: ResearchCommunicationGC34.
DeGood, R. & Chartrathi, K. (1991) Comparativeanalysis of tests work studying
factors influencing pressuresdeveloped in vented deflagrations. Journal of Loss
Preventionin theProcessIndustries,4,297-304.
DeHaan, J. D., Crowhurst, D., Hoare, D., Bensilum, M. & Shipp, M. P. (2001)
Deflagration involving stratified heavier than air vapor/air mixtures. Fire Safety
Journal, 36,693-710,
Donat, C. (1977) Pressurerelief as used in explosion protection. Journal of Loss
Preventionin theProcessIndustries,11,87-92,
Dorofeev,S. B., Bezmelnitsin,A. V. & Sidorov, V. P. (1995) Transition to detonation
in ventedbydrogen-airexplosions.Combustionand Flame, 103,243-246.
Eckhoff, R. K. (1991) Dust explosionsin the processindustries,Oxford, Butterworth-
Heinemann.
Ellis, d. C. 0. C. (1928) Flame movementin gaseousexplosiovemixtures. Fuel, 7,5-
12.
Fairweather,M., Hargrave,0. K., Ibrahim, S. S. & Walker, D. G. (1999) Studies of
premixedflamepropagationin explosiontubes.Combustionand Flame, 116,504-518.
Ferrara,G., Willacy, SK, Phylaktou,H.N., Andrews, G.E., Benedetto,A.D., Salzano,
E., Russo,G. (2008)Venting of gasexplosion throughrelief ducts: Interactionbetween
internal and external explosion. Journal of HazardousMaterials, In press accepted
manuscript
 224

Ferrara, G., Benedetto, A. D., Salzano, E. & Russo, G. (2006) CFD analysis of gas

explosions vented through relief pipes (article in press). Journal of Hazardous

Materials.
Ferrara, G., Willacy, S. K., Phylaktou, H. N., Andrews, G. E., Benedetto, A. D. &
Mkpadi, M. C. (2005) Duct vented propane-air explosions with central and rear
ignition. IAFSS.
Ferrara, G., Willacy, S. K., Phylaktou, H. N., Andrews, G. E., Benedetto, A. D. &
Salzano, E. (2005) Venting of premixed gas explosions with a relief pipe of the same

area as the vent. Proceedings of the European Combustion Meeting 2005.

Franzini, J. B. & Finnemore, E. J. (1994) Fluid mechanics with engineering
applications, New York, The McGraw-Hill Companies Inc.
Gardner, C. L. (1998) Turbulent combustion in obstacle-acceleratedgas explosions-the
influence of scale.Thesis. Department of Fuel and Energy. University of Leeds.
Gostintsev, Y. A., Istratov, A. G. & Shulenin, V. (1989) Self-similar propagation of a
free turbulent flame in mixed gas mixtures. Combustion, Explosion & Shock Waves
(Translatedfrom Fizika Goreniyai Vzryva, Vol 24, No. 5, pp. 63-70, Sept 1988), 563-
569.
Groff, E. G. (1982) The cellular nature of confined spherical propane-air flames.
Combustion and Flame, 4 8,51-62.
Gu, X J., Emerson, D. R. & Bradley, D. (2003) Modes of reaction front propagation
from hot spots. Combustion and Flame, 133,63-74.
Harris, G. F. P. (1967) The effect of vessel size and degree of turbulence on gas phase

explosion pressuresin closed vessels. Combustion and Flame, 11,17-25.

Harris, G. F. P. & Briscoe, P. G. (1967) The venting of pentane vapour-air explosions in

a large vessel, Combustion and Flame, 11,329-338.

Harris, R. J. (1983) The investigation and control of gas explosions in buildings and
heating plant, London, E&F N Spon Ltd.
Harris, R. J. & Wickens, M. J. (1989) Understanding vapour cloud explosions-an

experimental study. 55th Autumn Meeting, The Institution of Engineers.

Harrison, A. J. & Eyre, J. A. (1987) External explosions as a result of explosion

venting. Combustion, Science and Technology, 52,91-106.

Hey, M. (1991) Pressure relief of dust explosions through large diameter ducts 'and

effects of changing the position of ignition source. Journal of Loss Prevention in the
Process Industries, 4,217-222.
 225

Iida, N., Kawaguchi,0. & Sato,G. T. (1985)Premixedflame propagationinto a narrow

channel at a high speed,part i: Flame behavioursin the channel. Combustionand
Flame,60,245.
Istratov, A. G. & Librovich, V. B. (1969) Computing the speed of the normal
propagationof a flame in a mixture of hydrogenand chlorine,Astronautica Acla, 14,
453-467.
Kasmani,R. M., Andrews,G. E., Phylaktou,H. N. & Willacy, S. K. (2007a)Influence
of static burst pressureand ignition position on duct-vented gas explosions. 51h
InternationalSeminaron Fire and ExplosionHazards.Edinburgh.
Kasmani,R. M., Andrews,G. E., Phylaktou,H. N. & Willacy, S. K. (2007b)Vented gas
explosionin a cylindrical vesselwith a ventduct. EuropeanCombustionMeeting 2007
Chania,Crete.
Kasmani, R. M., Willacy, S. K., Phylaktou,H. N. & Andrews, G. E. (2006) Self-
acceleratinggasflamesin large ventedexplosionsthat are not accountedfor in current
vent design. Proceedings of the 2nd International Conference on Safety and
Environmentin ProcessIndustry.Naples,Italy.
Kordylewski,W. & Wach, J. (1988) Influenceof ducting on explosionpressure:Small
scaleexperiments.Combustionand Flame,71,51-61.
Kumar, R. K., Bowles, E. M. & Mintz, K. J. (1992) Large-scale dust explosion
experimentsto determinethe effects of scaling on explosion parameters.Combustion
andFlame, 89,320-332.
Kumar, R. K., Dewit, W. A. & Greig, D. R. (1989) Vented explosion of hydrogen-air

mixturesin a largevolume. Combustion,Scienceand Technology,66,251-266.

Kumar, R. K., Skraba,T. & Greig, D. R. (1987) Vented combustionof hydrogen-air
mixturesinlarge volumes.Nuclear Engineeringand Design,99,305-315.
Lee, J. H. S. & Guirao, C. M. (1982) Pressuredevelopmentin closed and vented
vessels.PlantlOperationalprogress,1,75-85.
Lees, F. P. (1996) Loss prevention in the processindustries: Hazard identification,
assessmentandcontrol, London, Butterworth-Heinemann,
Liebman,I., Corry, J. & Perlee,H. E. (1970)Flamepropagationin layeredmethane/air
system.Combustion,Scienceand Technology,1,257-267.
Lunn, G. (1992)Dust explosionpreventionand protection,Part I- venting, Institute of
ChemicalEngineers.
 226

Lunn, G., Crowhurst,D. & Hey, M. (1988) The effect of vent ducts on the reduced
explosionpressuresof venteddust explosions.Journal of Loss Prevention in Process
Industries,1,182-196,
Maisey, H. R. (1965) Gaseousand dust explosion venting- Part 1. Chemical and
ProcessEngineering,527-535.
Markstein,G. H. (1964)Nonsteadyflame propagation,New York, Macmillan.
McCann, D. P. J., Thomas,G. 0. & Edwards,D. H. (1985) Gasdynamicsof vented
explosionspart i: Experimentalstudies.Combustionand Flame,59,233-250.
Moen, 1.0., Bjerketvedt, D., Jenssen,A. & Thibault, P. A. (1985) Transition to
detonationin a large fuel-air cloud (brief communication).Combustionand Flame, 61,
285-291.
Moen, 1.0., Lee, J. H. S., Hjertager,B. H., Fuhre,K. & Eckhoff, R. K. (1982) Pressure
developmentdue to turbulentflame propagationin large-scalemethane-airexplosions.
Combustionand Flame,47,31-52.
Molkov, V., Dobashi,R., Suzuki, M. & Hirano, T. (2000) Venting of deflagrations:
Hydrocarbon-airand hydrogen-airsystems.Journal of Loss Prevention in the Process
Industries,13,397-409.
Molkov, V. V. (1994)Venting of deflagrations:Dynamic of the processin systemswith

a duct and receiver.Proceedingsof the 4th International Symposiumon Fire Safety

Science,1245-1254,
Molkov, V. V. (1995) Theoreticalgeneralizationof international experimentaldata on
ventedgasexplosiondynamics.Physicsof Combustionand Explosions,165-181.
Molkov, V. V. (1999)Explosionsafetyengineering:NFPA 68 andimproved vent sizing
technology. Interflam 1999, Proceedings of the 81h International Fire Science
Conference.Edinburgh.
Molkov, V. V. (2001) Unified correlationsfor vent sizing of enclosuresat atmospheric
and elevatedpressures.Journal of Loss Preventionin the ProcessIndustries, 14,567-
574.
Molkov, V. V., Baratov,A. & Korolchenko,A. (1993) Dynamics of gas explosionsin
ventedvessels:Review and progress.Progressin Astronautics and Aeronautics, 154,
117-131.
Molkov, V. V., Dobashi, R., Suzuki, M. & Hirano, T. (1999) Modelling
of vented
hydrogen-airdeflagrationsand correlationsfor vent sizing, Journal Loss Prevention
of
in the Processindustries, 12,147-156.
 227

Molkov, V. V., Grigorash,A. V., Eber,R. M. & Makarov,D. V. (2004)Ventedgaseous

deflagrations: Modelling of hinged inertial vent covers. Journal of Hazardous

materials,Al 16,1-10.
Molkov, V. V., Grigorash,A. V., Eber, R. M., Tarnanini,F. & Dobashi, R. (2004)
Vented gaseousdeflagrationswith inertial vent covers-State-of-the-artand progress.
ProcessSafetyProgress,23,29-36.
Molkov, V. V., Kokolchenko,A. & Alexandrov, S. (1997)Venting of deflagrationsin
buildings and equipment:Universalcorrelation.Fire SafetyScience-proceedings
of the
5th InternationalSymposium,1249-1260.
Molkov, V. V. & Nekrasov,V. P. (1981) Dynamicsof gas combustionin a constant
volume in the presenceof exhaust.Balashika. Translatedfrom Fizika Goreniya i
Vzryva,17,17-24.
Munday, G. (1963)The calculationof venting areasfor pressurerelief of explosionsin
vessel. Proceedingsof Symposiumon Chemical Process Hazards with Special
Referenceto Plant Design Series No-15. Manchester,The Institution of chemical
Engineers.
Munson,B. R., Young,D. F. & Okiishi, T. H. (2006)Fundamentalsoffluid mechanics:
Fifth edition,JohnWiley andSonsInc.
Nagy, J. & Verakis, H. C. (1983) Developmentand control of dust explosions,New
York, MarcelDekkerInc, New York andBasel.
NFPA68 (2002) Nfpa 68: Guide for venting of deflagrations:2002. National Fire
ProtectionAssociation.
Ng, H, D. & Lee,J. H. L. (2007)Commentson explosionproblemsfor hydrogensafety.
Jounal of Lossof Preventionin ProcessSafety.
Pasman,H. J., Groothuizen,T, M. & Gooijer, H. d. (1974) Design of pressurerelief
vents. Loss Preventionand Safety Promotion in the ProcessIndustries: Edited by
C.H.Buschmann,185-189.
Phylaktou, H. N. & Andrews, G. E. (1991) Gas explosionsin long closed vessels.
CombustionScienceand Technology,77,27-39.
Phylaktou,H. N., Andrews, G. E. & Iferath, P. (1990) Fast flame speedsand rate of
pressurerise in the initial period of gas explosionsin large l/d cylindrical enclosures.
Journal of Lossof Preventionin ProcessSafety,30355-364.
Phylaktou,H. N., Foley, M. & Andrews, G. E. (1993)Explosionsin a tube with a 90"
bend.Journal of Lossof Preventionin ProcessSafety,6,21-29.
 228

Ponizy, B. & Leyer, J. C. (1999a)Flame dynamicsin a ventedvesselconnectedto a

duct: I. Mechanismof vessel-ductinteraction.Combustionand Flame, 116,259-271.
Ponizy, B. & Leyer, J. C. (1999b)Flame dynamicsin a ventedvesselconnectedto a
duct: 2.1nfluenceof ignition site, membranerupture and turbulence.Combustionand
Flame, 116,272-281.
Ponizy, B. & Veyssiere,B. (2000) Mitigation of explosions in a vented vessel
to
connected a duct. Combustion, Science
and Technologyý158,167-182.
Pritchard, D. K., Allsopp, J. A. & Eaton, G. T. (1995) Gas explosion venting in
elongatedenclosures.1-12.
Rasbash,D. J. (1986) Quantificationof explosion parametersfor combustible fuel-air
mixtures.Fire SafetyJournal, 111113-125.
Rasbash,D. J., Drysdale,D. D. & Kemp, N. (1976) Design of an explosion relief
systemfor a building handlingliquefied fuel gases.I. Chem.E.SymposiumSeriesN6.47,
145-156.
Razus, D. M. & Krause, U. (2001) Comparison of empirical and semi-empirical
calculationmethodsfor ventingof gasexplosion.Fire SafetyJournal, 36,1-23.
Runes,E. (1972)Explosionventing.Plant Operations& LossPrevention,6,63-7 1.
Russo,P. & Benedetto,A. D. (2007) Effects of a duct on the venting of explosions-

critical review. Trans1ChemE,Part B (ProcessSafetyand EnvironmentalProtection),

85(B1), 9-22.
Searby, G. & Quinard, J. (1990) Direct and indirect measurementsof Markstein

numbersin premixedflames.Combustionand Flame, 82,298-311.

Simpson, L. L. (1986) Equations for the VDI and Bartknecht nomograms. Plant
/OperationsProgress,8,49-51.
Siwek, R. (1996) Explosion venting technology. Journal of Loss Prevention in the
ProcessIndustries,9,81-90.
Solberg, D. M., Pappas,J. A. & Skramstad, E. (1980) Observations of flame
instabilities in large scale vented gas explosions. 181hhiternational Symposiumon
Combustion.Universityof Waterloo,Canada,Det NorskeVeritas.
Starke,R. & Roth, P. (1986) An experimentalinvestigationof flame behaviour during
cylindrical vesselexplosions,Combustionand Flame, 66,249-259.
Swift, 1. (1983) Gaseouscombustionventing- a simplified approach,4th International
Symposiumon LossPreventionand Safetypromotion in the ProcessIndustries,3, F21-
F37.
 229

practice. Plantl0perational progress, 3,

Swift, 1. (1984) Venting deflagrations-theory
89-93.
Swift, 1. & Epstein, A (1987) Performanceof low pressure explosion vents.
Plantl0perationsprogress,6,98-105.
Tamanini,F. (2000)Partialvolumedeflagrationcharacteristicsof explosionsin layered
fuel/air mixtures.InternationalSeminaron Fire and ExplosionHazardsof Substances.
Tamanini, F. & Fisher, M. (2003) Mixed-mode venting of dust explosions. 41h
InternationalSeminaron Fire and ExplosionHazards.Londonderry.
Thorne,P. F., Rogowski,Z., W& Field, P. (1983)Performanceof low inertia explosion
reliefs fitted to a 22m3cubicalchamber.IN ENGINNER, T. 1.0. C. (Ed.) 4th Int. Sym.
on Loss Preventionand SafetyPromotion in the Process Industries (Series no.82).
Harrogate,England,The Instituteof ChemicalEngineers.
Tseng, L. K., Ismail, M. A. & Faeth, G. M. (1993) Laminar burning velocities and
Marksteinnumbersof hydrocarbon/airflame. Combustionand Flame,95,410-426.
Ural, E. A. (1993) A simplified methodfor predicting the effect of ducts connectedto

explosionvents.Journal of LossPreventionin the ProcessIndustries,6,3- 10.

Wingerden,C. J. M. & Zeeuwen,J. P. (1983a)On the role of acousticallydriven flame
instabilitiesin ventedgasexplosionsand their elimination. Combustionand Flame, 51,
109-111.
Wingerden, C. J. M, & Zeeuwen,J. P. (1983b) Venting of gas explosions in large
rooms. 4tn InternationalSymposiumon Loss Preventionand Safety Promotion in the
ProcessIndustries.
Yao, C. (1974)Explosion ventingof low-strengthequipmentand structures.Journal of
LossPreventionin theProcessIndustries,8,1 -9.
Zalosh, R. (1995)Review of gasdeflagrationventing models. I" InternationalFire and
ExplosionHazardsof Substances andVenting of Deflagrations,79-87.
Zalosh, R. G. (1980) Gas explosion tests in room-size vented enclosures. Loss
Prevention,13,98-110.
APPENDIX
 "30

'F',Ihle A. IASIOl'tllC JILIHISIILd

experiniciltal clataI'or vented gas cxl)los"()"','.
-2

Methane-air

Colic. V A Vessel I./D Ignition 1), 13jed A, expt Sources

vol 11, (m3) (m2) shape ratio (barg) (bar,-) (m2)

o)
10.4 0. 1 93 Rcc I. MIR' 0 0.0 1 0.0 1', ý R /idwdl. K 4. VC111cm, cl i"
-1 .

10.4 0.2 1.93 Rec 1.5 Centre 0.03 0.02 0.03)15 RX I. /alosh. K 6. Vent covei is

alumill Itill)
10.4 0.2 1.93 Rec 1.5 Centre WO 0.00 0.0 172 K.( i. /alosh. KII. Vcnt cover is

alumillitini

10.4 0.2 1.9-1 Rec 1.5 Centre 0.0 0.21 0.0007S K.( J.Aalosh. K 28. Vent cover is

almilinimll

10.4 0.2 2.01 Rec I Centre 0.03 0.0 0.054 R.G. /alosh. K 5.8. Vent cover is

aluillillillill

10.4 0.2 2.01 Kec I Centre 0.01 0.06 0.031 R.Gzalosh. K 10. Vent cover is

alLill1illiLlIll
10.4 0.2 2.01 Rec I Centre 0.03 0.21 0.01(1 R.G./Aosh. K 20. Vent cover is

altilifillimil
10.4 0.2 2.01 Rec I Centre 0.03 0.69 0.0077 R.G./alosh. K-41. Vent co\,cr is

alunlinium
10.4 0.2 2.08 Rec I Centre 0.0.1 0.01 0.045 R.C. /alosh. K- 5.8. Will cover is

aluillinillill
10.4 0.2 2.08 Rec I Cclitrc 0.03 0.04 0.0". 3 R.O./alosh. K 8. Vent covvr is

10.4 0.2 2.08 Rec I Centre 0.03 0.14 0.015 luj. /alosh. K 18. Vent cover i's

allimilliulli

10.4 0.2 2.08 Rec I CCIIII-C 0.0.1 0-48 0.0001) KAi. /alosh. K 18. Vent covci Is

.1111111inium

NR 0.4 Rec I CCIIII-C 0 0.043 0.3 Cubbage and S111111londs1: [oil

3.25
1-cliefsill bo\ ovcli.
w(lb/11") activallon pressuic
NR 0.4 3.25 Rcc I Centre 0.002 0.08 0.3 Ctibbagc and Sillullonds 1: lop

relicl's ill box oven. \V(1hAl

activation I-)I.Csstll.c

NR 0.4 3.25 Rec I Centre 0.00.) 0.1 0.1, Clibbagc and Sililmonds 1: 101)

rcliel'S Ill box oven. WOWIC)

 231

I I activation pressure
NR 0.4 3.25 Rec I Centre 0.004 0.13 0.3 Cubbage and Simmonds 1: Top

reliefs in box oven. W(lb/fl)=

activation pressure
10.4 0.6 4.38 Rec 1.5 Centre 0.03 0.03 0.0652 R.G.Zalosh. K=6. Vent cover is

alurninium
10.4 0.6 4.38 Rec 1.5 Centre 0.03 0.05 0.049 R.G.Zalosli. K-8. Vent cover is

aluminium
I I
10.4 0.6 4.38 Rec 1.5 Centre 0.03 0.76 0.036 R.G.Zalosli. K=l 1. Vent cover is

aluminium
10.4 0.6 4.38 Rec 1.5 Centre 0.03 0.36 0.0098 R.G.Zalosh. K=40. Vent cover is

aluminium
9.5 1.0 5.5 Cylinder 1.554 Cent e 0.16 1.0 0.1 P.F.Thome et at. Polyethylene
l30um, K=4,0. lgnition at the
center of rear wall
9.5 1.0 5,5 Cylinder 1.554 Centre 0.32 2.0 0.05 S.Chippett
9.5 1.0 6.0 Cubic I Centre O'l 0.2 0.36 Bartknecht(Abb.2.223a)
9.5 1.0 6.0 Cubic I Centre 0.2 0.2 0.36 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 Cubic I Centre 0.2 0.4 0.36 Bartknecht(Abb.2,223b)
9.5 1.0 6.0 Cubic I Centre 0.5 0.64 0.36 Razus et at
9.5 1.0 6.0 Cubic I Centre 0.5 0.8 0,36 Bartknecht(Abb.2.223b)
9.5 1.0 6.0 Cubic I Centre 0.8 0.9 0.36 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 Cubic I Centre 1.0 1.2 0.36 Bartknecht(Abb.2.223b)
9.5 1.0 6.0 Cubic I Centre 1.0 1.3 0.36 Bartknecht(Abb.2.223(a))
9.5 Lo 6.0 Cubic I Centre 1.5 1.7 0.36 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 1 Cubic I Centre 2.0 2.4 0.36 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 Cubic I Centre 0.1 0.43 0.3 Razusetal
9.5 1.0 6.1 Cylinder 1,55 Centre 0.18 0.5 0.3 Paper as vent cover. One peak

pressureAst peak is taken as Pred-

PasmanjU et a]
9.5 1.0 6.1 Cylinder 1.55 Centre 0.20 0.1 0.3 Mylar as vent cover. Double peak

pressureAst peak is taken as P.O.

PasmanjU ct at
9.5 1.0 6.0 Cubic I Centre 0.2 0.77 0.3 Razuset at
9.5 1.0 6,0 Cubic I Centre 0.5 0.59 0.3 Razuset at
9.5 1.0 6.0 Cubic I Centre 0.5 0-5 0.26 Bartkneclit(Abb. 2.224(b))
9.5 1.0 6.0 Cubic I Centre 0.2 0.5 0.22 Bartknecht(Abb.2,224(b))
9.5 1.0 6.0 Cubic I Centre 0.1 0.5 0.2 Bartknecht(Abb,2.224(b))
9.5 1.0 6.0 Cubic I Centre 0.84 0.2- Razuset at
.1
 232

9.5 1.0 6.1 Cylinder 1.55 Centre 0.18 1.2 0.2 Paper as vent cover. One peak

pressure.Pasmaii,I-I.J et at
9.5 1.0 6.1 Cylinder 1.55 Centre 0.20 0.1 0.2 Mylar as vent cover. Double peak

pressure.2nd peak is taken as P,,d-

PasmanjU et a[

9.5 1.0 6.0 Cubic I Centre 0.2 1.48 0.2 Razuset al

9.5 1.0 6.0 Cubic I Centre 0.5 1.11 0.2 Razuset at

9.5 1.0 6.0 Cubic I Centre 0.1 1.0 0.16 Bartkneclit(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 0.2 1.7 0.16 Bartkneclit(Abb.2.223b)

9.5 1.0 6.0 Cubic I Centre 0.2 1.8 0.16 Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 0.5 1.2 0.16 same as above
9.5 1.0 6.0 Cubic I Centre 0.5 1.4 0,16 Razuset al

9.5 1.0 6.0 Cubic I Centre 0.8 1.3 0.16 Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 1.0 1.4 0.16 l Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 1.0 115 0,16 Bartknecht(Abb.2.223b)

9.5 1.0 6.0 Cubic I Centre 1.5 2.0 0.16 Bartknecht(Abb.2.223(a)),

9.5 1.0 6.0 Cubic I Centre 2.0 2.5 0.16 Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 0.5 1 2.0 0.12 l Bartknecht(Abb.2.224(b))

9.5 LO 6.0 Cubic I Centre 0.1 1.4 0.1 Razuset at

9.5 1.0 6.0 Cubic I Centre 0.2 2.0 0.1 Bartknecht(Abb.2.224(b))

9.5 1.0 6.0 Cubic I Centre 0.2 2.19 0.1 Razuset at

9.5 1.0 6.1 Cylinder 1.55 Centre 0.20 1.0 0.1 Mylar as vent cover. Double peak
I I I pressure.Pasman,113 et al
9.5 1.0 6.0 Cubic I Centre 0.5 1.95 0.1 Razuset al

9.5 1.0 6.0 Cubic I Centre 0.1 2.0 0,08 Bartknecht(Abb.2.224(b))

9.5 1.0 6.1 Cylinder 1.55 Centre 0.20 2.0 0.05 Mylar as vent cover.One peak

pressure.Pasman,ti. 3 et al
9.5 1.0 6.0 Cubic I Centre 0.1 1.8 0.04 Bartknecht(Abb.2.223b)

9.5 1.0 6.0 Cubic I Centre 0.2 2.6 0.04 Bartkneclit(Abb. 2,223(a))

9.5 1.0 6.0 Cubic I Centre 0.2 2.8 0.04 Bartknecht(Abb.2.223b)

9.5 1.0 6.0 Cubic I Centre 0.5 3.2 0.04 Bartknecht(Abb.2.223b)

9.5 1.0 6.0 Cubic I Centre 0.5 3.4 0.04 Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre 0.8 3.4 0.04 Bartknecht(Abb.2.223(a))

9.5 1.0 6.0 Cubic I Centre LO 2.8 0.04 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 Cubic I Centre 1.5 2.6 0.04 Bartknecht(Abb.2.223(a))
9.5 1.0 6.0 Cubic I Centre 2.0 3.0 0,04 1Bartkneclit(Abb. 2.223(a))
1.9 7.4 Sphere I Centre 2.9 9.3 10.0993 S,Chippett.
Sphere I Centre 3.5 11.2 10.0993 1S.Ehippett, Turbulent with 2 fan.
 233

10 1.9 7.4 Sphere I Centre 4.5 10.2 0.0993 S.Chippett,

10 1.9 7.4 Sphere I Centre 6.0 10.0 0.0993 S.Chippett.

10 1.9 7.4 Sphere I Centre 2.8 19.8 0.0324 S.Chippett.

10 1.9 7.4 Sphere I Centre 3.0 20.4 0.0324 S.Chippett.

10 1.9 7.4 Sphere I Centre 5.2 21.1 0.0324 S.Chippett.

NR 2.8 12.46 Rec 1.8 Centre 0 0.014 1.14 Cubbage and Simmonds 1: Top

reliefs in box oven. W(lb/ft2)=

activation pressure
NR 2.8 12.46 Rec 1.8 Centre 0.002 0.034 1.14 Cubbage and Simmonds 1: Top
W(lb/fl2)=
reliefs in box oven.
activation pressure
NR 2.8 12.46 Rec 1.8 Centre 0.005 0.066 1.14 Cubbage and Simmonds I: Top
W(lb/ft2)=
reliefs in box oven.
activation pressure
10 3.8 11.8 Sphere I Centre 2.8 15.8 0.0993 S.Chippett. Turbulent with 2 fan.

10 3.8 11,8 Sphere I Centre 2.9 15.4 0.0993 S.Chippett.

10 3.8 11.8 Sphere I Centre 3.5 17.9 0.0993 S.Chippett.

10.5 11.2 31.54 Rec 0.9 Centre 0.03 0.5 2.25 R.G.Zalosh.P2 is taken.
I Aluminium foil as a membrane

10.5 11.2 31.54 Rec 0.9 Centre 0.03 0.3 0.77 R.G.Zalosh.P2 is taken.
I Aluminium foil asa membrane
10 22.0 47.11 Cubic I Centre 0.012 0,016 7.85 P.F.Thorne et al. Polyethylene
130gm:K=I
10 22.0 47.11 Cubic I Centre 0.03 0.04 3,41 P.F.Thorne et al. Polyethylene
I 130gm;K=2.3
10 22.0 47.11 Cubic I Centre 0.05 0.02 3.41 P.F.Thorne et al. Polyethylene
130gm; K=2,3: Ignition at the
centreof front wall
10 22.0 47.11 Cubic I Centre 0.05 0.07 3.41 P.F.Thorne et al, Polyethylene
130i,im; K=2.3: Ignition at tile
centreof rear walI
10 22.0 47.11 Cubic I Centre 0.05 0.06 3.41 P.F.Thorne et al. Fibrcboard

I 12.5iim,K=2.3
10 22.0 47.11 Cubic I I Centre 0.05 0.16 3,41 FI.F.Thorne et al. Polyester50lim;
K-2.3
10 22.0 47.11 Cubic I Centre 0.04 0.07 1.96 P.F.Thorne et al. Polyethylene
130gm;K-4.0
10 22.0 47.11 Cubic I Centre 0.05 0.11 1.96 P.F.Thorne et al, Polyethylene
I I I I 1 130^ K-4.0* Ignition at tile
 234

I I
centreof rearwall
10 22T 47.11 Cubic I Centre 0.09 0.6 0.98 P.F.Thorne et aL Fibreboard
12.5pm;K=8.0
10.0 26.6 56.1 Rec I Centre 0.02 0.07 0.50 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec I Centre 0.06 0.11 0.50 Buckland(G.A, Lunn book)
10.0 26.6 56.1 1 Rec I Centre 0.06 0.05 0.50 Buckland(G.A, Lunn book)
10.0 26.C 56.1 Rec I Centre 0.07 0.06 0.50 Buckland(G.A. Lunnbook)
10.0 26.6 56.1 Rec I Centre 0.04 0.11 0.40 Buckland(G.A. Lunn book)
10.0 26.6 56.1 Rec I Centre 0.07 0.11 0.40 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec I Centre 0.02 0.10 0.25 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec 3 Centre 0.07 0.07 0.52 Buckland(G.A. Lunnbook)
10.0 26.6 56.1 Rec 4 Centre 0.08 0.13 0.63 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec 2 Centre 0.09 0.07 0.40 Buckland(G.A.Lunn book)
10.0 26.6 156.1 Rec I Centre 0.09 0.22 0.20 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec I Centre 0.12 0.22 0.20 Buckland(G.A.Lunn book)
10.0 26T 56,1 Rec - I Centre 0.007 0.08 0.13 Buckland(G.A.Lunn book)
10.0 26.6 56.1 Rec I Centre 0.04 0.06 0.13 Buckland(G.A,Lunn book)
9.5 30.0 57.93 Cubic I Centre 0.1 0.1 3.9 Bartknecht(Abb.
2.222).F= nx
I DN 1000
9.5 30.0 57.93 Cubic I Centre 0.1 0.2 3.4 Bartknecht(Abb.
2.222).F= nx
DN 1000
9.5 30.0 57.93 Cubic I Centre 0.1 0.4 2.5 Bartknecht(Abb.
2.222).F=n x
I I DN800
9.5 30.0 57.93 Cubic I Centre 0.1 0.5 2.0 2.222).F=n
Bartknecht(Abb. x
DN800
9.5 30.0 57.93 Cubic I Centre 0.1 1.0 L5 2.222).F=n
Bartknecht(Abb. x
I DN800
9.5 30.0 57.93 Cubic I Centre 0.1 0.9 1.4 Bartknecht(Abb.
2.222),F=n x
DN600
9.5 30.0 57.93 Cubic I Centre 0.1 1.4 1.0 Bartknecht(Abb.
2.222).F=n x
I DN800
9.5 30.0 57.93 Cubic I Centre 0.1 1.88 0.85 Bartknecht(Abb.
2.222).F=n x
DN600
9.5 30.0 57.93 Cubic I Centre 0.1 2.2 0.8 Bartkneclit(Abb.
2.222).F= 11 x
I DN 1000
9.5 30.0 57.93 Cubic I Centre 0.1 2.6 0.5 Bartkneclit(Abb.2.222).F=n x
DN800
9.5 30.0 57.93 Cubic I Centre 0.1 3.7 0.3 Bartknccht(Abb.
2.222).F-n x
I I I I I I I I IDN600
 10.5 34.0 66.12 Rec 2.7 Centre 0.0 1 1.0 3 RA Ualosli. pl i.,, taken.

Altiminium 10il &, o membrofic

10.5 34.0 66.12 Rec 2.7 Centre 0-031 2.2 1 R.( ;,zalo."JI.P1 I,, taken.
-
Altumnium 1'()ilas a Ilicillbralle

10.5 34.0 66.12 Rec 2.7 Centre 0.011 1 25 KA I./alosh. P2 11.1 taken.
.3
Aluminlum 1'()il Illembralle

10.5 34.0 66.12 Rec 2.7 Centre 2.25 R.(,. ZaIosh.P2 is taken.
Aluminlum I'Oil Inell1hralic
(Tentre - 7 -
)-031 1.6 5 R.G.Zalosh. A 0.91-m dianicler
10.5 34.0 06.12 Rec 2T
and 2.7m long tank and an arraý ol
2.5cm pipc, suspended froll,

Ceiling is placcd ill thc ClIclosuivs.

0-03) 1.7 I
1.86 R.G./alosh. 211d Vent palicl hC111
10.5 34.0 66.12 Rec 2.7 (7entrc
open 2SIko

Propane/air

Conc. V A, Vessel CID Fition

Tdr Pv Pred A,,expt Source
(Vol %) (M) (M)
shape ratio (barg) (barg) (M)
0.021) 0.0 1(0) 1 1)(1111c
cl '11

5.0 0.03 0.61 Vessel 1.54 Center 4.8 10.9 0.055 Cousin &, Cottons( 1051

5.0 0.03 0.61 Vessel 1.54 Center 29.0 32.1 0.055 l Cousin & Cottons( 1951

5.0 0.03 0.61 Vessel 1.54 Center 33.1 331.1 0.055 Cousin & Cottons( 1951).

5.0 0.03 0.61 Vessel 1.54 Center 16.4 20.7 0.055 Initial pressure is 4-5 psig(.'). 10

5.0 0.03 0.61 Vessel 1.54 Center 16.6 23.4 0.021 COLISin& Cottons( 1051

5.0 0.03 0.61 Vessel 1.54 Center 4.8 15.2 0.021 Cousin & Cottons( 1951

0.03 0.61 Vessel 1.54 Center 24.8 30.3 0.021 Cousin & Cottons( 1()51
-5.0
5.0 0.03 0.61 Vessel --1.54 -Uentci* 4.9 22.1 0.01 1 Cousin & Cottons( 11)51
ýý 54 Center 18.6 29 0.011 COLISin& Cottons( 1951
5.0 0.03 0.61 Vessel
.
5.0 0.03 0.61 Vessel 1.54 Center 20 33.1 0.011 cousin & cottons( 1951

5.0 0.085 1.134 Tank 2.3 Centre 1.5 2.8 0.0417 Cousin A, Cottons( 1951

5.0 0.085 1.134 Tank 2.3 Centre 9,0 8.0 0.0417 Cousin Colton.,"( 1951)

5.0 0.085 1.134 Tank 2.3 Centre 10.6 11.9 0.0417 Comill Cottons( 1951
5.0 0.085 1.1'14 Tank 2.3 Centre 0.0 0.1 0.0408 Cousin & Cottons( 1951)

5.0 0.085 1.134 Tank Centi 1.8 5.3 0.0 101 Cousin & Uottons(l 951
- -1 I .1
5.0 0.095 1.1'34 Tank 2.1 Centre 7.0 8.9 0.0 161 Cousin & Cottons( I )ý I
T-ank 2.3 Centre 12.7 13.3 0.0 161 CoUsin & Cottons( 195 1
 236

5.0 0.085 1.1341 Tani_ _233 -Centre 0.0 1.1 0,0135 Cousin& Cottons(1951)

5.0 0.085 1.134 Tank 2.3 Centre 1.6 6.5 0.0080 Cousin & Cottons(I 95 1)

5.0 0.085 1.134 Tank 2.3 Centre 5.9 9.6 0.0080 Cousin & Cottons(I 95 1)

5.0 0.085 1.134 Tank 2.3 Centre 11.9 13.7 0.0080 Cousin & Cottons(I 95 1)

5.0 0.085 1.1341 Tank 2.3 Centre 0.0 2.0 0,0060 Cousin & Cottons(] 95 1)

5.0 0.085 1.1341 Tank 2.3 Centre 0.0 2.7 0.0036 Cousin & Cottons(l 95 1)

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.021 0.07 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.028 0.07 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 1.55 0.0525 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.048 0.0525 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 1.32 0.0438 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.0276 0.035 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.344 0.0314 R.G.Zalosh

4.60 0.17 1.9 Rec 1.5 Centre 0.03 0.028 0.0314 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.037 0.0314 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.28 0.0314 R.G.Zalosh

5.20 0.17 1.9 Rec 1.5 Centre 0.03 0.66 0.0314 R.G.Zalosh
1
5.20 0.17 1.9 Rec 1.5 Centre 0.03 1.61 0,0263 R.G.Zalosh

6.00 0.17 1.9 Rec 1.5 Centre 0.03 0.03 0.0175 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 1 0.03 0.055 10.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.26 0.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.05 0.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.74 0.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.05 10.0525 R.G.Zalos

5.20 0.18- 2.0 Rec I Centre 0.03 0.052 0.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.052 0.0525 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.083 0.035 R.G.Zalosh

5.20 0.18 2.0 Rec I Centre 0.03 0.57 0.035 R.G.Zalosh

4.80 0.65 2.9 Cylinder 1.198 Centre 0.86 2.75 0.0993 P.F.Thome et al

4.80 0.65 2.9 Cylinder 1.198 Centre 0.86 3.14 0.073 P.F.Tliorne et al

4.80 0.65 2.9 Cylinder 1.198 Centre 0.72 5.59 0.0324 P.F.Thorne et al

4.80 0.65 2.9 Cylinder 1.198 Centre 0.65 5.22 0.0182 P.F.Thome et al

4.80 0.65 2.9 Cylinder 1,198 Centre 0.65 5.31 0.0182 P.F.Thome et al

4.5 0.77 5.0 Cubic I Center 0.049 0.01 0.52 Yao

4.5 0.77 5.0 Cubic I Center 0.0686 0.021 0.52 Yao
4.5 0.77 5.0 Cubic I Center 0.0196 0.0103 0.52 Yao
4.5 0.77 5.0 Cubic I Center 0.0924 0.0171 0.29 Yao
4.5 0.77 5.0 Cubic I Center 0.063 0.035 0.29 Ya
 237

4.5 0.77 5.0 Cubic I Center 0.06 0.03 0.20 Yao

4.5 0.77' 5.0 Cubic I Center 0.1096 0.175 0.16 Yao

4.5 0.77 5.0 Cubic I Center 0.1078 0.1848 0.16 Yao

5.00 0.77 5.0 Cubic I Centre 0.0 0.05 0.29 Razus et al

4.5 0.79 5.7 Cylinder I Center 0.0938 0.0595 0.85 Yao

4.5 0.79 6.7 Cylinder I Center 0.043 0.0203 0.85 Yao

4.5 0.79' 4.7 Cylinder I Center 0.04 0.056 0.59 Yao

4.5 0.79 6.7 Cylinder I Center 0.0 0.0175 0.53 Yao

4.5 0.79 6.7 Cylinder I Center 0.0 0.0286 0.30 Yao

4.5 0.79 6.7 Cylinder I Center 0.0 0.04 0.21 Yao

4.5 0.79 4.7 Cylinder I Center 0.108 0.08 0.06 Yao

4.00 1.0 6.0 Cubic I Centre 0.1 0.14 0.6 Razus et al

4.00 1.0 6.0 Cubic I Centre 0.1 0.32 0.4 Razus et al

4.02 1.0 1.0 Cubic I Centre 0.1 0.28 0.38 Bartknecht data(Abb.2.219)

5.00 1.0 4.8 Sphere I Centre 0.2 0.3 0.355 Donat

5.00 1.0 4.8 Sphere I Centre 0.1 0.3 0.31 C.Donat(l 973)

5.00 1.0 4.8 Sphere I Centre 0.22 0.52 0.301 C.Donat(1973)

5.00 1.0 4.8 Sphere I Centre 0.2 0.5 0.3 Donat

4.02 1.0 1.0 Cubic I Centre 0.5 0.5 0.3 Bartknecht(Abb.2.224(b))

4.02 1.0 1.0 Cubic I Centre 0.2 0.5 0.259 Bartknecht(Abb.2.224(b))

5.00 1.0 4.8 Sphere I Centre 0.1 0.5 0.25 C.Donat(1973)

4.02 1.0 1.0 Cubic I Centre 0.1 0.5 0.24 Bartknecht(Abb.2.224(b))

4.02 1.0 1.0 Cubic I Centre 0.1 0.55 0.23 Bartknecht data(Abb.2.219)

5.00 1.0 4.8 Sphere -1 Centre 0.2 1 0.21 Donat

4.00 1.0 6.0 Cubic I Centre 0.1 1 0.2 Razus ct al

5.00 1.0 4.8 Sphere I Centre 0.5 1 0.195 C. Donat(1973)

5.00 1.0 4.8 Sphere I Centre 0.52 1.02 0.195 Donat

4.02 1.0 1.0 Cubic I Centre 0.1 0.92 0.18 Bartknecht data(Abb.2.219)

5.00 1.0 4.8 Sphere I Centre 0.1 1 0.15 C. Donat(1973)

5.00 1.0 4.8 Sphere I Centre 0.2 1.5 0.145 Donat

5.00 1.0 4.8 Sphere I Centre 0.22 1.52 0.145 C. Donat(1973)

4.02 1.0 1.0 Cubic I Centre 0.5 2.0 0.1 Bartknecht(Abb.2.224(b))

5.00 1.0 4.8 Sphere I Centre 0.5 1.5 0.12 C. Donat(1973)

4.02 1.0 1.0 Cubic I Centre 0.2 2.0 0.1 Bartkneelit(Abb. 2.224(b))
4.02 1.0 1.0 Cubic I Centre 0.1 1.9 0.1 Bartknecht data(Abb.2.219)
4.02 1.0 1.0 Cubic I Centre 0.1 2.0 0.1 Bartkneclit(Abb. 2.224(b))
5.00 1.0 4.8 Sphere I Centre 0.2 2 0.095 Donat
5.00 1.0 4.8 Sphere I Centre ni C. Donat(l 973)
 238

5.00 1.0 4.8 Sphere I Centre 0.5 2 0.085 C.Donat(1973)

5.00 1,0 4.8 Sphere I Centre 0.2 2.5 0.07 Donat

5.00 1.0 4.8 Sphere I Centre 0.5 2.5 0.065 C.Donat(1973)

5.00 1.0 4.8 Sphere I Centre 0.52 2.52 1 0.065 C.Donat(I 973)

Centre 0.1 2.4 0.055 Bartknecht ýata(Abb.2.219)

4.02 1.0 6.0 Cubic I
5.00 1.0 4.8 Sphere I Centre 0.1 2 0,045 C.Donat(1973)

5.00 1.7 6.9 Sphere I Centre 0.52 1.02 0.308 1larris and Briscoc

5.00 1.7 6.9 Sphere I Centre 0.22 1.52 0.156 Harris and Briscoe

5.00 1.7 6.9 Sphere I Centre 0.22 0.52 0.139 Harris and Briscoe

5.00 1.7 6.9 Sphere I Centre 0.52 2.52 0.139 Harris and Briscoe

4.02 2.0 9.5 Cubic I Centre 0.1 0.29 0.33 Bartknecht(Abb.2.224(b))

4.02 2.0 9.5 Cubic I Centre 0.1 0.55 0.26 Bartknecht(Abb.2.224(b))

4.02 2.0 9.5 Cubic I Centre 0.1 1 0.17 Bartknectit(Abb.2.224(b))

4.02 2.0 9.5 Cubic I Centre 0.1 2.1 0.12 Bartknecht(Abb.2.224(b))

5.00 2.6 11.91 Rec 2.3 Centre 0.096 0.199 0.56 R.DeGood/K.Chatratlii: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Bottom 0.103 0.159 0.56 R.DeGood/K.Chatratlii: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.1 0.235 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Ree 2.3 Centre 0.1 0,314 0.56 R.DeGood/K. Chatrathi: Study of
factors affecting Pred.
I
5.00 2.6 11.91 Rec 2.3 Centre 0.086 0.185 0.56 R.DeGood7K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.086 0.311 0.56 R,DeGood/K. Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.086 0.385 0.56 R.DeGood/K. Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Bottom 0.086 1.007 0.56 R.DeGood/K.Chatrathi- Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.086 0.215 0.56 R.DeGood/K. Cliatrathi: Study of
Ifactors affecting Pred.
5.00 2.6 11.91 Ree 2.3 Centre 0.93 0.241 0.56 R.DeGood/K. Cliatrathi: Study of
factors affecting Pred.
5.00 2.6 11.91 Rec 2.3 Centre 0.9 0.172 0.56 R.DeGood/K. Chatrathi: Study of
I factors affecting Pred.
5.00 2.6 11.91 Rec 2.3 Centre 0.9 0.16 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.
5.00 2.6 11.91 Rec 2.3 Centre 0.9 0.144 0.56 1R.DeGood/K.Cliatratiii: Study of
-
 239

factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.262 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0,103 0.26 0.56 R.DeGood/K.Cjlatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.156 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Cen re 0.11 0.246 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.
I
5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.334 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0,103 0.325 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.29 0.56 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.441 0.56 R.DeGood/K.Chatrathi: Study of
I. factors affecting Pred.

5.00 2.6 11.91 Rec 2.3 Centre 0.103 0.282 0.54 R.DeGood/K.Chatrathi: Study of
factors affecting Pred.
5.00 2.6 11.91 Rec 2.3 Centre 0.1 0.3 0.54 R.DeGood/K.Chatrathi: Study of
I I I factors affecting Pred.

5.00 10.0 22.5 Sphere I Centre 0.2 0.3 2.18 C.Donat(1973)

5.00 10.0 22.5 Sphere I Centre 0.1 0.3 2A C.Donat(l 973)
5.00 10.0 22.5 Sphere I Centre 0.2 1 0.5 1.7 C.Donat(l 973)
5.00 10.0 22.5 Sphere I Centre 0.1 0.5 1.5 C.Donat(l 973)
5.00 10.0 22.5 Sphere I Centre 0.5 1 1.21 C.Donat(I 973)
5.00 10.0 22.5 Sphere I Centre 0.2 1 1.07 C.Donat(I 973)
5.00 10.0 22.5 Sphere I Centre 0.1 1 0.8 C.Donat(1973)
5.00 10.0 22.5 Sphere I Centre 0.5 1.5 0.79 1C.Donat(I 973)
5.00 10.0 22.5 Sphere I Centre 0.2 1.5 0.74 C.Donat(I 973)
5.00 10.0 22.5 Sphere I Centre 0,5 2 0.56 C.Donat(1973)
5.00 10.0 22.5 Sphere I Centre 0.1 1.5 0.55 C.Donat(1973)
4.02 10.0 4.6 Cubic I Centre 0.1 0.24 0.49 Bartkneclit(Abb. 2.219)
5.00 10.0 22.5 Sphere I Centre 0.2 2 0.47 C.Donat(I 973)
5.00 10.0 22.5 Sphere I Centre 0.5 2.5 0,42 C.Donat(1973)
5.00 10.0 22.5 Sphere I Centre OA 2 0.4 C,Donat(1973)
4.02 10.0 4.6 Cubic I Centre 0.1 0.32 0.4 Bartknecht(Abb.2.219)
4.02 10.0 4.6 Cubic I Centre 0.1 0.39 0.38 Bartknecht(Abb.2.219)
10,0 22.5 Sphere I Centre 0.2 2.5 0.37 C.Donat(1973)
 240

4.02 10.0 4.6 Cubic I I Centre 0.1 0.52 0.32 Bartknectit(Abb.2,219)

5.00 10.0 22.5 Sphere I Centre 0.1 2.5 0.3 C.Donat(l 973)

4.02 10.0 4.6 Cubic I Centre 0.1 0.69 0.28 Bartkneclit(Abb.2.219)

4.02 10.0 4.6 Cubic I Centre 0.1 0.9 0.2 Bartknecht(Abb.2.219)
4.02 10.0 4.6 Cubic I Centre 0.1 2.8 0.1 Bartknecht(Abb.2.219)

4.02 10.0 27.8 Cubic I Centre 0.1 2.9 2.219)

0.052 Bartknecht(Abb.
4.05 ll'o 28.3 Cylinder 1.25 Rear 0.05 0.09 1.36 P.F.Thorneetal
4.3 22.0 47.11 Cubic I Centre 0.02 0.025 7.85 P.F.Thomeetal
4.3 22.0 47.111 Cubi I Centre 0.04 0.2 3.41 P.F.Thomeet al
4.3 22.0 47.111 Cubic I Centre 0.09 0.19 3.41 P.F.Thorneetal
4.3 22.0 47.111 Cubic I Centre 0.04 0.2 1.96 P.F.Thomeetal
4.3 22,0 47,11 Cubic I Centre 0.05 0.25 1.96 P.F.Thorneetal
4.3 22.0 47.11 Cubic I Centre 0.2 0.55 1.96 P.F.Thorneet al
4.5 25.0 53.9 Rec I Center 0.0 0.025 7.70 Bromma(Sweden)
6.0 25,0 53.9 Rec I Center 0.0 0.01 6.20 Bromma(Sweden)
5.00 30.0 46.7 Sphere I Centre 0.5 0.5 3.48 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.1 0.3 3 C.Donat(I973)
5.00 30,0 46.7 Sphere I Centre 0.2 0.3 2.7 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.1 0.5 2.25 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.2 0.5 2.03 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.5 1 1.88 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.1 1 1.5 C.Donat(l973)
5.00 30,0 46.7 Sphere I Centre 0.2 1 1.35 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0.5 1.5 1.35 C.Donat(l973)
5.00 30.0 46.7 Sphere I Centre 0.2 1.5 0.97 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 05
0'
0.5 2 0.97 C,Donat(1973)
5.00 30.0 46.7 Sphere I Centre 0.1 1.5 0.9 C.Donat(l973)
5.00 30.0 46.7 Sphere I Centre 0.2 2 0.72 C.Donat(I973)
5.00 30.0 46.7 Sphere I Centre 0
0.5 2.5 0.72 C.Donat(I973)
5.00 130.0 46.7 Sphere I Centre 0,1 2 0.6 C.Donat(I973)
5.00 30,0 46.7 Sphere I Centre 0.2 2.5 0.53 C,Donat(1973)
5.00 30.0146.7 Sphere I Centre 0.1 2.5 0.45 C.Donat(I973)
4,02 30.01 9.7 Cubic I Centre 0.1 0.2 0.4 Bartkneclit(Abb.
2.219)
4.02 30.0 9.7 Cubic I Centre 0.1 0.6 0.2 Bartkneclit(Abb.
2.219)
4.02 30T ' 9.7 Cubic I Centre 0.1 0.75 0.175 Bartknecht(Abb.
2.219)
4.02 30.0 9.7 Cubic I Centre 0.1 1.1 -0.17 Bartknecht(Abb.
2.219)
402 30.0 9.7 1 Cubic I Centre 0.1 1.7 0.1 Bartkneelit(Abb.
2.219)
30.0 ý7.91- Cubic I Centre 0.) 2.1 0.069 Blrtknecht(Abb.
2,219)
 241

4.45 30.4 61.4 Rec 0.363 Centre 0.4 0.7 0.58 P.F.Thome et at
4.3 35.0 67.4 Rec 1.14 Center 0.1 0.18 2.1 Solberg et al(1980)
4.3 35.0 67.4 Rec 1.14 Front 0.1 0.19 2.1 Solberg et a1(]980)

4.3 35.0 67.4 1 Rec 1.14 Rear 0.1 0.1 2.1 Solberg et al(1980)

4.3 35.0 67.4 Rec L 14 Front 0.1 0.3 1.6 Solberg et al(I 980)

4.3 35.0 67.4 Rec 1.14 Rear 0.1 0.18 1.6 Solberg et al(1980)
4.3 35.0 67.4 Rec 1.14 Center 0.1 1.05 1.1 ISolberg et al(1980)

4.3 35.0 67.4 Re 1,14 Front 0.1 0.71 1.1 Solberg et al(1980)

4.3 35.0 67.4 Rec 1.14 Rear 0.1 0.72 1.1 Solberg et al(I 980)

5.00 35.0 67.4 Rec 1.415 Near 0.0 0.75 1 P.F.Thorne ct at

I vent
5.00 35.0 67.4 Rec 1.415 Centre 0.0 1.37 1 P.F,Thome et at
4.3 35.0 67.4 Rec 1.14 Center 0.1 1.55 0.53 Solberg et al(I 980)
4.3 35.0 67.4 Rec 1.14 Front 0.1 1.68 0.53 Solberg et a](1980)
6.0 40.0 73.7 Rec 'I Center 0.0137 0.0299 8.48 Bromma(Sweden)
6.0 40.0 73.7 Rec 'I Center 0,0196 0.0295 8.48 Bromma(Sweden)
5.0 60.0 74.2 Sphere I Centre 0.5 0.5 5.22 C.Donat(l 973)
5.0 60.0 74.2 Sphere I Centre 0.1 0.3 4.8 C.Donat(1973)
5.0 60.0 74.2 Sphere I Centre 0.2 0.5 3.76 C.Donat(I 973)
5.0 60.0 74.2 Sphere I Centre 0.1 0.5 3 IC.Donat(1973)
5.0 60.0 74.2 Sphere I Centre 0.5 1 2.22 C.Donat(I 973)
5.0 60.0 74.2 Sphere I Centre 0.1 1 1.2 C.Donat(I 973)

4.0 60.0 92.0 Cubic I Centre 03 1.5 0.69 Bartknecht(Abb.2.219)

4.0 60.0 92.0 Cubic I Centre 0.1 0.4 0.22 Bartkneclit(Abb,2.219)
4.0 60.0 92.0 Cubic I Centre 0.1 0.6 0.18 Bartknecht(Abb.2.219)
4.0 60.0 92.0 Cubic I Centre 0.1 1 0.14 Bartknecht(Abb.2.219)
6.0 70.0 107.0 Rec I center 0.0 0.0785 12.31 Bromma(Sweden)
6.0 70.0 107.01 Rec I Center 0.0196 0.03481 12.31 Bromma(Sweden)
6.0 200,0 215.5 Rec I Center 0.055 0.06 1 30.81 Bromma(Sweden)
6.0 200.0 215.5 Rec I Center 0.0294 0.033 30.81 Bromnia(Sweden)
6,0 200.0 215.5 Rec I Center 0.0098 0.018 30.81 Bromma(Sweden)
6.0 200.0 215.5 Rec I Center 0.029 0.0343 30.81 Bromrna(Sweden)
6.0 200.0 215.51 Rec I Center 0.0549 0.0588 30.81 Bromina(Sweden)
6.0 200.0 215.5 Rec I Center 0.049 0.0637 18.19 Brornma(Swcden)
4.0 203.8 218.2 Rec 1.951 centre 0.0 0.03 21.58 P.F.Thorne et at
4.0 1203.81218.2 Rec 1.951 Centre 0.0 0.06 17.3 P.F.Thorne et at
5.2 0.17 Rec 0.3 Center.] 0.036 0.046 0.05 Zalosh
.-1.9 -
 I ')

5.2 0.17 1.9 Rec 03 1 Center 0.036 0.03 80 0.05 zalosh

5.2 0.17 1.9 Rec 1 ('Unter 0.036 0.05 0.05 Z; Ilosll

5.2 0.17 1.() Rec 0,3 Center 0.03 6 0.069 0.oý Z; iiosli

5.2 0.17 1.1) Rec 0.3 Center 0.036 0. OW 0.09 zalos11

1
5.2 0.17 1.1) Roc 0. 0.036 0.0552 0.08 zah. ýII
-1
5.2 0.17 1.9 Rec 0.' Center 0.036
1 0.0621 0.08 zillosli

5.2 0.17 1.9 Rec 0.3, Center 0,0-36 0.0481 0A0 zalos11
1 1 1
5.2 0.17 1.9 Rec 0.3 Center 0.036 0.0552 0.10 Allosh

5.2 0.18 2.0 Rec 1 Celltel. 0.036 0.10-3, 0. OS zalosil

5.2 0A8 2.0 Rec 1 Cellter 0.0-36 ()AM) 0,05 Zalosll

0.18 2.0 Rec 1 (, ellter 0.036 0. W)65 0,05 zalosh

-5.2
5.2 0.18 2.0 Rec 1 Center 0.036 0.0965 0.05 Zalosh

5.2 0.18 2.0 Rec 1 Center 0.036 0.112 0.05 zalos11

5.2 0.18 2.0 Rec 1 Center 0.036 0.112 0.05 Allos11

5.2 0,19 2.0 Rec 1 Center 0.036 0.104 0.05 zalosh

5.2 0.18 2.0 Rec 1 Center 0.036 0.0896 0.03 zalosil

5.2 0.18 2.0 Rec 2 Center 0.0-16 0.105 0.033 Zalosil

5.0 81.00 117.9 Rec 1 Center 0.014 0.049 18.72 flowards and Karabinis.

5.0 91.00 117.9 Rec 2 Cellter 0.0252 0.0462 18.72 Howards and Karahinis.

5.0 81.00 117.9 Rec 3 Cellter 0.0735 0.1008 18.72 1lowards and Karahinis.

5.0 81.00 117,9 Rec 4 Center 0.0351 0.0469 18.72 Howards and Karahinis.

5.0 91.00 117.9 Rec 5 Celltel, 0.0 16 0.035 19.72 1lowards and Karabinis,
_3
5.0 81.00 117.9 Rec 6 ('eiltet- 0.0136 0.043 18.72 1Imards and Karahinis-

5.0 81.00 117.9 Rec 7 Center 0.0189 0.0679 18.72 1lowards and Karabinis.

5.0 81.00 117.9 Rec 8 Center 0.014 0.021 18.72 1lowards and Karahinis.

Town gas/ air

IgIlitiOn 1)," ----------

(Vol ýýo) (III on) sliallu ratio (barg) (bill -g) (III)

0,000S 011 0 uhh;wc ""Immond"

25 0.23 2.25 Rec I I Center 0.0008 0.12 0.17 Cubbage& sillillionds
25 0.23 2.25 Rec I Center 0.001 0.13 0.37 Cubbagc & Simmonds
25 0.23 2.25 Rec I Center 0.001-1 0.14 0,3T T'llbbage & sillillionds
25 0.23 2.25 Rec I Center 0.0012 0.11 0..1 Cubbage& similloilds
25 0.23 2.25 Rec I Center 0.00096 0.14 0-1 Cubbage& silliniond.."
2-5 0,23 2.25 Rec I Center 0.0017 0.18 0.1 Cubbilge & sillillionds
L 25 0.23 2.25 Re Center 0.0017 0,19 0.3 1Cubbagc & S1111111ollds
 243

25 0.23 2.25 Rec I Center 0.00072 0.16 0.21 Cubbage& Simmonds

25 0.23 2.25 Rec I Center 0.001 0.22 0.21 Cubbage& Simmonds

25 0.23 2.25 Rec I Center 0.0013 0.26 0.21 Cubbage& Simmonds

25 0.23 2.25 Rec I Cenleý 1.00019 0.08 0.092 l Cubbage& Simmonds

25 0.23 2.25 Rec I Center 0.00072 0.19 0.092 Cubbage& Simmonds

25 0.23 2.25 Rec I Center 0.00086 0.24 0.092 Cubbage& Simmonds

10 1.47 7.76 Rec 4.48 Center 0.0005 0.01 0.84 Cubbage& Simmonds

15 1.47 7.76 Rec 4.48 Center 0.0005 0.029 0.84 Cubbage& Simmonds

20 1.47 7.76 Rec 4.48 Center 0.0005 0.069 0.84 Cubbage& Simmonds

25 1.47 7.76 Rec 4.48 Center 0.0005 0.096 0.84 Cubbage& Simmonds

30 1.47 7.76 Rec 4.48 Center 0.0005 0.041 0.84 Cubbage& Simmonds

35 1.47 7.76 Rec 4.48 Center 0.0005 0.022 0.84 Cubbage& Simmonds

25 1.47 7.76 Rec 4.48 Center 0.0008 0.069 0.84 Cubbage& Simmonds

25 1.47 7.76 Rec 4.48 Center 0.001 0.131 0.84 Cubbage& Simmonds

25 1.47 7.76 Rec 4.48 Center 0.003 0.248 0.84 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0012 0.076 1.24 Cubbage&, Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0013 0.083 1.24 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0026 0.141 1.24 Cubbage& Simmonds

25 2.8 11.92 , Rec 1.3 Center 0.0005 1 0.041 1.14 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0014 0.086 1.14 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.002 0.124 1.14 Cubbage& Simmonds

25 2.8 11.92 Rec 1,3 Center 0,0026 0.15 1.14 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0006 0.076 0.84 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.001 0,103 0.84 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.002 0.14 0.84 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0006 0.034 0.62 Cubbage& Simmonds

25 2.8 11.92 Rec 1,3 Center 0.0007 0.09 0.62 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0014 0.17 0.62 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0007 0.103 0.47 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.001 0.124 0.47 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.001 0.131 0.47 Cubbage& Simmonds

25 2.8 11.92 Rec 1.3 Center 0.0012 0.152 0.47 Cubbage&, Simmonds

25 1.0 6.00 Cubic I Center 0.1 0.5 0.26 Bartkneclit(Abb. 2.224(b))

25 1.0 6,00 Cubic I Center 0.2 0.5 0.28 Bartkneclit(Abb. 2.224(b))
25 1.0 6.00 Cubic I Center 0.5 0.5 0.32 Bartknecht(Abb.2.224(b))
25 1.0 6.001 Cubic I Center 0.1 2.0 0.12 Bartkncclit(Abb. 2.224(b))
25 1.0 6.00 Cubic I Center 0.2 1 2.0 0.13 lBartkneclit(Abb. 2,224(b))
2.5 1 1.0 6.00 Cubic I I Center 0.5 2.0 0.15 Bailknecht(Abb. 2.224(b))
 `44

I lydrogen/air

N)-nc. V A, vessel L/D Ignition P', Sources

2)
(Vol (M"') (m ratio (barg) (barg) (111")
shape
I ).i Inilial P 4S
O.OT 0.00-1 \ csscl
-1 .:
psig (3.1 harg)
1.ý4 Centre 1 1.2 25.5 0.011 COUSill & COHOlls. llliti; ll P . 1ý
40 0,03 0.004 Ves..
wl
psig (3.1 bilru)

Vessel 1.54 Centre 29.5 0.011 S. Initial

Cousi 11 & ('01101I. P -1ý
40 0.03 0.004
psig barg)

Vesse I -T-54 Centre 4.8 11-1 0.011 Cousin A, Cottons. Initial 1' 4S
40 0.01 0.004
psi (3.1 barg)

0.03 0.004 Vessel -ý-.54 -77ct-itrc 11.7 22.1 0.021 Cousill & Cottoll..". Initial 11 45
40
psh-,(.). I barvo

40 0.03 0.004 Vessel 1.54 Centre 27.6 0.021 COLISill & COU011S.llliliýd P -15

psig barg)

Vessel 1.54 C'ciltrc 4.8 11.7 O.Osý Cousin & Cottons. Initial 1) 4S
40 0.03 0.004
psig (3.1 barg)

40 0.03 0.004 Vessel -Fý4- -Z7- I1 10.5 0.055 Cousin & Cottons. hillial P 45
elltrC
psig (3.1 barg)
1 1
40 0.0.1 0.004 Vessel Lý4 Centre 22.8 2S.4 0.055 Cousin & Cottons. Initial P 4S

psig (3.1 barg)

Lý4 Centre 28.3 28.8 0.055 S. Illitlill

COU.Sill & ('01101I. P 45
40 0.03 0.004 Vessel
psig (3.1 baro)

40 0.085 1.134 Tank 2-3) (7entre 0 3.1 0.14 Cousin & Cottons. Initial P0

pSig

40 Tank 2.3 Centre 0 2.0 0.26 Cotisin Cottom. Initial 11 0

0.08.5 1.13,4
psig

40 0.085 1.134 Tank 2 Centre 0 1.2 0.72 Cousin Cotlolls, Initial 11 0

_3
psig
-c -11t I. o ).2 0.008 Cousin C011011". Initial PI
40 0.08S 1.134 '1ank c re
1)siLý

40 0.085 1.134 Tank 2.3 CCIIII-C 5.() 11 0.008 Cousill Cottons 1111tialP I.S
.
1)Sig

0.085 1.134 Tank 2.3 Centre 11.9 12.o 0.008 Cousin A-, ('011011s. inilial PIS
140

O.W 1.114 '1ank 2.3 Centre 1.8 TO 0.0101 Cousin & Cottoll". Initial PIS
-10
psig
40 0.085 1.134 '1ank 2.3 Centre 7 10.3 0.0161 COLISill & COttOllS. 11116illP 1.5
 245

I psig
40 0,085 1.134 Tank 2.3 Centre 11 11.9 0.0161 Cousin & Cottons. Initial P= 15

psig
40 0.085 1.134 Tank 3.3 Centre 1.5 4.6 0.0417 Cousin & Cottons. Initial P= 15

psig
40 0.085 1.134 Tank 4.3 Centre 6.5 8.2 0.0417 Cousin & Cottons. Initial P= 15
I psig
40 0.085 1.134 Tank 5.3 Centre 10.6 11.8 0.0417 Cousin & Cottons. Initial P= 15

Ipsig
40 0.22 0.5 Drum 1.41 Centre 0 4.5 0.09 Cousin & Cottons. Initial P= 0

psig
40 0.22 0.5 Drum 1.41 Centre 0 2.8 0.23 Cousin & Cottons. Initial P= 0

psig
40 0.22 0.5 Drum 1.41 Centre 0 1.5 0.43 Cousin & Cottons. Initial P= 0

psig
40 0.4 5.8 Pipe 22.1 Centre 0 4 0.05 Cousin & Cottons. Initial P= 0

psig
40 0.4 5.8 Pipe 22.1 Centre 0 2.9 0.09 Cousin & Cottons. Initial P= 0

psig
40 0.4 5.8 Pipe 22.1 Centre 0 2.6 0.2 Cousin & Cottons. Initial P= 0
I psig
40 0.4 5.8 Pipe 22.1 Centre 0 1.3 0.69 Cousin &, Cottons. Initial P= 0

1 1 psig
NR 0.95 4.673 Cylinder 2 Centre 0.075 1.3 0.2 1Razuset a]
NR 0.95 4.673 Cylinder 2 Centre 0.135 0.4 0.3 Razuset at
20 1 6 Cubic I Centre 0.1 0.5 0.34 Bartknecht(Abb.2.224(b))
20 1 6 Cubic I Centre 0.1 2 0.159 Bartknecht(Abb.2.224(b))
20 1 6 Cubic I Centre 0.2 0.5 0.36 Bartknecht(Abb.2.224(b))
20 1 6 Cubic I Centre 0.2 2 0.16 Bartknecht(Abb.2,224(b))
20 1 61 Cubic I Centre 0.5 2 0.18 Bartknecht(Abb.2.224(b))
10 6.37 16.6 Sphere I Centre 0.1 0.3 0.64 R.K. Kumar et.a).() 989)
10 6.37 16.6 Sphere I Centre 0.1 0.9 0.2 R.K. Kurnar et.al,(l 989)
10 6.37 16.6 Sphere I Centre 0.1. 1.7 0.071 R.K. Kumar et.al,(I 989)
20 6.37 16.6 Sphere I Centre 0.1 3.4 -U4 R.K. Kumar et.al.(1989)
20 6.37 16.6 Sphere I Centre 0.1 4.2 0.2 R.K. Kumar et,al,(I 989)
20 6.37 16.6 Sphere I Centre 0.1 4.8 0.071 R.K. Kumar et.al.(] 989)
12 6.37 16.6 Sphere II Centre 0.1 1 0,64 R.K. Kumar et.al.(I 9 89)
12 637 16.6 Sphere I Cent e 0.1 2 0.2- R.K. Kumar et.al.(] 989)
12 Sphere I Centre 0.1 2.5 0.071 R.K. Kurnar et.al.(I 9 89)
 246

14 6.37- 16.6 Sphere I Centre 0.1 1,6 0.64 R.K. Kumaret. al.(1989)

14 6.37' 16.6 Sphere I Centre 0.1 4.9 0.2 R.K. Kurnar et.al.(l 989)

14 6.37 16.6 Sphere I Centre 0.1 5.1 0.071 R.K. Kumar et.al.(1989)

16 6.37 16.6 Sphere I Centre 0.1 2.2 0.64 R.K. Kumar et.al.(1989)

16 6.37 16.6 Sphere I Centre 0.1 3.4 0,2 R.K. Kumar et.al.(l 989)

16 6.37 16.6 Sphere I Centre 0.1 3.6 0.071 R.K. Kurnar et.al.(1989)

18 6.37 16.6 Sphere I Centre 0.1 3 0.64 R.K. Kumar et.al.(1989)

18 6.37 16.6 Sphere I Centre 0.1 3.9 0.2 R.K. Kumar et.al.(] 989)

18 6,371 16.6 1 Sphere I "ntre 0.1 5.1 0.071 R.K. Kumar et.al.(l 989)

Acetone/air

Conc. V As Vessel P, red A,.,,, Sources

p
2)
(vol %) (m3) (m2) shape (barg) (barg) t(m

3.0 0.008 -0.19 Sphere 0 1.17 0.00013 Proceedingof I st & 2nd Sym. On Comb.Inst.

3.0 0.008 0.19 Sphere 0 0.48 0.00051 Proceedingof I st & 2nd Sym. On Comb.Inst.

3.0 0.008 0.19 Sphere 0 0.14 0.002 Proceedingof I st & 2nd Sym. On Comb. Inst.

3.0 0,008 0.19 Sphere 0 0.034 0.0082 Proceedingof I st & 2nd Sym. On Comb.Inst.

3.0 0.008 0.19 Sphere 0 0.034 0.018 Proceedingof I st & 2nd Sym. On Comb.Inst.

3.0 0.008 0.19 Sphere 0 0.034 0.031 Proceedingof I st & 2nd Sym. On Comb. Inst.

3.0 0.008 0.19 Sphere 0 0.034 0.0016 Proceedingof I st & 2nd Sym. On Comb, Inst.

3.0 0.008 0.19 Sphere 0 0.034 0.0036 Proceedingof I st & 2nd Sym. On Comb. Inst.

4.0 0.008 0.19 Sphere 0 2.76 0.00013 Proceedingof Ist & 2nd Sym. Oil Comb.Inst.

4.0 0.008 0.19 Sphere 0 1.38 0.00051 Proceedingof I st & 2nd Sym. On Comb.Inst.

4.0 0.008 0.19 Sphere 0 0.48 0.002 Proceedingof I st & 2nd Sym. On Comb.Inst.

4.0 1 0.19
0.008 Sphere 0 0.17 0.0082 Proceedingof I st & 2nd Sym. On Comb.Inst.

4.0 8 0.19
0.008 Sphere 0,034 0.018 Proceedingof I st & 2nd Syrn. On Comb.Inst.

4.0 8 0.19
0.008 Sphere 10 0.034 0.031 Proceedingof I st & 2nd Sym. On CombAnst.

4.0 0.008 0.19 Sphere 0 0.17 0.0016 Proceedingof I st & 2nd Sym. On Comb.I nst.

4.0 0.008 0.19 Sphere 0 0.034 0.0036 Proceedingof I st & 2nd Sym. On Comb.Inst.

5.0 0.008 0.19 Sphere- -0 4.48 0.00013 Proceedingof I st & 2nd Sym. On Comb.Inst.

5.0 8 0.19
0.008 Sphere 0 3.1 0.00051 Proceeding of I st & 2nd Sym. On Comb.Inst.

5.0 0.008 0.19 Sphere 0 0.83 0.002 Proceedingof Ist & 2nd Sym. On Cornb.lnst.

5.0 0.008 0.19 Sphere 0 0.28 0.0082 Proceeding of I st & 2nd Sym, On Comb.Inst.

5.0 0.008 0.19 Sphere 0 0.1 0.018 Proceeding of I st & 2nd Synn,On Comb.Inst.

5.0 0.008 0.19 Sphere 0 0.0344 0.031 Procecclingof I st & 2nd Syrn. On Comb-Inst.

5.0 0,008 0.19 Sphere 0 0.34 0.0016 Proceeding of I st & 2nd Sym. On Comb.Inst.
-
5.0 0.008 ý 0.19 1 Sphere 0 0.1 0.05367 Proceedingof I st & 2nd Sym. On Comb.Inst.
 247

5.5 0.008 0.19 Sphere 0 3.38 0.00051 Proceedingof I st & 2nd Sym. On Comb.1nst.

5.5 0.008 0.19 Sphere 0 0.97 0.002 Proceedingof I st & 2nd Sym. On Comb.Inst.

5.5 0.008 0.19 Sphere 0 0.31 0.0082 Proceedingof I st 2nd Sym. On CornbAnst,

5.5 10.008 0.19 1 Sphere 0 0.14 0-018 Proceedingof I st 2nd Sym. On Comb.1nst.
.
5.5 0.008 0.19 Sphere 0 0.034 0.031 Proceedingof Ist & 2nd Sym. On CombAnst.

5.5 0.008 0.19 Sphere 0 0.41 0.0016 ProceedingofI st & 2nd Sym. On Comb.1nst.
5.5 0,008 0.19 Sphere 0 0.14 0.0036 Proceedingof I st &, 2nd Syrn. On Comb.] nst.
4.907 0.022 0.374 Cylinder 0.100 0.680 0.680 Molkov et al: Venting of Deflagration: IIC-air
I and H2-air system
4.907 2.000 7.677 Cylinder 0.150 2.270 2.270 Molkov et a]: Venting of Deflagration: HC-air

and H2-air system

4.907 0.022 0.374 Cylinder 0.200 0.500 0.500 Molkov et al: Venting of Deflagration: IIC-air

and H2-air system

4.907 0.022 0.374 Cylinder 0.230 I 0.480 0.480 Molkov et al: Venting of Deflagration: HC-air

and H2-air system

4.907 2.000 7.677 Cylinder 0,750 3.170 3.170 Molkov et al: Venting of Deflagration: HC-air

I and H2-air system

4.907 0.760 4.027 Cylinder 1.900 3.300 3.300 Molkov et a]: Venting of Deflagration: HC-air

and H2-air system

5.5 8.05 0.319 Cylinder 0.000 0.070 3.653 G.A Lunn
1
5.5 8.05610.319 Cylinder 0.000 0.100 2.734 G.A Lunn

5.5 8.056 0.3191 Cylinder 0.000 0.175 1.818 G.A Lunn

5.5 8.056 0.3191 Cylinder 0.000 0.315 0.909 G.A Lunn

 ."8

Tablc A. 2 List ol'publislied experimental data t'()]-vellwd gas explosion mth a pi-C.
"Clicc

Pillic.
()I'CILICI
A, P" Prod Ignition Prod Souces
Fuel L D Ud (M)
(M) (M) (M) (bar) (bar) without

vent (bar)

I'l-opillic ill] 0. () 10 0.0 1 00, (Io 1,01 lo I Ild \oI Polil"\ mid I -\cl

111casillud

4.0(ýi, 0.6 0.021 28.6 0.000140 0.00366 1.0 1 1-11d Not POlll/\ and LCN,
cl-
-1.18
measurccl
0.6 0.036 16.7 0.001017 0.003166 1.01 2.2 8 End Not Pollizý and Leyer

mcasurcd
1.1 0.016 68.8 0.000201 0.00366 1.01 2.81 Fnd Not Poili/ý and kcýer

111casill-cd
1.1 0.021 52.4 0.000146 0.00366 1.01 2.40 I.,nd Not Polli/) and Leyel.

measured
1.1 0.036 30.6 0.001017 0.00366 1.01 2,93 I-Ind Not Poliizý and Leyer

measured
2.6 0.016 162.5 0.000201 0.00366 1.01 2.93 Fild Not Polliz), and Leyer

measured
2.6 0.021 123.8 0.00031460.00366 1.01 2.56 F,nd Not Volliz) and Lever

measured
2.6 0.036 72.2 0.0010170.00366 1.01 2.4 Fnd Not Ponizýand Lcyer
111casill-cd
1.7 0.036 47.2 0.001017 0.00366 1.01 3.02 Central Not and Leyer
11011i/ý

nicasured
1.7 0.036 47.2 0.001017 0.00366 1.31 3.17 Centrill Not llollizý and Leyer

measured
1.7 0.036 47.2 0.001017 0.00366 1.92 3.67 Central Not Pollin and I, CNef,

illeasurcd
1.7 0.036 47.2 0,001017 0.00360 33 1 4.38 Cciltral Not Polii/ý and Leyel.

IIIC,tStll*CLI
1.7 0.036 47.2 0.001017 0.00360 1.01 1:11d Not 11onizyand I,cvef-
-1.77

1.7 0.036 47.2 0.001017 0.00306 1.33 2.99 1md Not Poni/y and I.c\,cr

111castil-cd
1.7 0.036 47.2 0.0010 17 0.00166 1.84 2.82 1 Ild Not Pollin and 1,c\ cl
 249

Propane/air
5.0% 1 0.8446 1.2 0.56 2.6 1 11 1.2 Central 1.212 DeGood a
.
Chartrathi
2 0.8446 2.4 0.56 2.6 1 11 1.32 Central 1,212 DeGood a
.
Chartrath!
3 0.8446 3.6 0.56 2.6 1 11 1.4 Central 1.212 DeGood a
.
Chartrathi
3 0.8446 3.6 0.56 2.6 1 11 2.02 Bottorn 1.16 DeGood a
.
Chartrathi

Acetone/air
5.0% 25 0.5 50.0 0.19625 10 1.11 5.11 Central Not Molkov (1993)
measured
25 0.5 50.0 0.19625 10 1.06 3.81 Central Not MoIkov (1993)
measured
4 0.2 20.0 0.0314 2 1.16 5.31 Central Not Molkov (1993)
measured
10 0.2 50.0 0.0314 2 1.16 6.21 Central Not Molkov (1993)
measured
10 0.38 26.3 0.113354 2 IA 3.16 Central Not Molkov (1993)
measured
1.83 0.05 36.6 0.001963 0.027 1.21 6.01 Central 1.7 Molkov (1993)
2.35 0.05 47.0 0.001963 0.027 1.26 5.41 Central 1.7 Molkov (1993)
2.35 0,05 47.0 0.001963 0.027 1.26 4.51 Central Not Molkov (1993)
measured
2.35 0.05 47.0 0.001963 0.027 2.66 2.91 Central Not Molkov (1993)
measured
1.83 0.05 36.6 0.001963 0.027 2,43 5.41 Central Not Molkov (1993)
measured

Methane/air
9.5% 0.11 0.1129 1.0 0.01 0.00564 1.01 1.11 Not 1.092 McCann(1985)
available
0.3 0.1129 2.7 0.01 0,00564 1.01 1.2 Not 1.092 McCann(1985)
available
0.52 0.1129 4.6 0.01 0.00564 1.01 1.18 Not 1.092 McCann(1985)
available