1

Explosion & Gas Release from LNG CarriersBy Gordon Mime, Senior Risk Analyst

Uoyd’s Registerof Shipping

INTRODUCTION

There is a growing trend throughout industrytowards using Liquthed Natural Gas (LNG) as apotentialsourceof energy.LNG is theliquefiedformof naturalgasand is predominatelymade up ofMethanewith the remainingsmallpercentagesmadeup of Ethane andPropane. This mix is transferredinto a liquid form to allow it to be transported by shipto its final destination.This is carriedout by loweringthe temperature of the gasuntil it is cold enough toform a liquid (462°C).

Despitethe excellent safety record of the shippingindustry, thereis still someconcern about whether somuchgaspresentsariskof major explosion. It is easyto acknowledgethatLNG doesburn,it doesform gasclouds,andit is extremelycold. However it has manyqualities which limit the consequencesof an LNGspill. This paperpresentsa high level overview ofhow LNG is likely to behaveduringareleasefrom aship. As such it puts in context the majorexplosiontheoryagainstwhat is knownin practice.

The approach taken looksat evaluatinga worst casescenarioof amajor releasefrom aLNG tanker. Suchreleasesareextremelyunlikely in comparisonto smallscalereleases. However examination of such largescalereleasespresent a worst casescenario.

This includesevaluation of the lossof containment,the formationof LNG pools, gas plumesandignitionhazards

The assessmentsof the consequencesare backedupby an evaluation of historical and experimentalevidence. This evidencehas beenpresentedto assistin the understanding of the nature of theconsequencesandto provide justification for certainconclusions.

GLOSSARY

Bund-A retainingwall or dyke designedto containliquid releasedusually as a resultof the failure of astoragetank.

Deflagration;- The low speed combustion of aflammablegascloud in which no damagingpressurewave is produced.

Detonation:- When thecombustion flamespeedin anignited gascloud increasesup to or above the speedof soundin thegas,a detonation is said to occur. Theflame front is directlycoupled to the pressureprofilewhich takes the form of a shock wave. Damagingoverpressures can occur which are transmittedoutside theregionof the gascloud.

DetonatiOnsgenerally occur in pipework or highlycongested regions of process plant. To date nodetonations have been produced in unconuinedmethanecloud tests.

DispersionModels:- Mathematicalmodelswhichareusedto predict the spread andshapeof a gas cloud.The models may be used to predict distances tospecified concentrationlevelswithin the cloud andhencegive concentrationcontourplots.

Emissive Power.- The heat flux measured at thesurfaceof a flame.

Flame Speech- The speed of propagation of acombustionflamethrougha gascloud. The fasterthespeed the higher the associated overpressureproduced. Flame speedsgreater than loom/s mayresult in damagingoverpressures.

Lower Flammable Limit (LFL):- The minimumquantity of flammablegas(usuallyexpressedas S byvolume) which when mixed with air will supportcombustion. For methaneair mixtures the LFL is 5%by volume, and for propane air mixturesthe LFL is2% by volume.

Overpressure:.for a pressurepulse(blastwave),thepressuredevelopedaboveatmosphericpressure.

Upper Flammable Limit (UFL):- The maximumquantity of flammablegas(usually expressedas S byvolume) which when mixed with air will supportcombustion.For methane-airmixturestheUFL is 15%by volume, and for propane-airmixtures the UFL is8% by volume.

Unconfined Vapour Cloud Explosion:- Anunconfinedvapour cloud explosion (UVCE) describesan explosionof a flammablevapour-airmixtureeitherin the openair or in partiallyconfinedcircumstancesdue to the presenceof buildings,structures,trees,etc.

CONFIDENTIAL

1k ‘1d 1~’mtnt~~1iflit In’. dUd tUt U ~iOU~ di tk t’.~ (I .iu tin i .tnd ott it I ltt\ d ‘~Rt ‘go.kr ot Shi~iping. L1n~dsol Sh rpt

~o is not ~ ~ ,~~ ~ , I~~ ~ auSd ~‘Lit ng ti tk~sub~tancc01

tlit’ artid~ Ans rc~lian~~on j~. thiii~ in thi~di (i~lv is jt tht it ri~k it 1k çvi st in wh~ivIlt”. on it~

2

HISTORICAL EVIDENCE

Modem LNG shipping has existed since 1965.Numbershave increaseduntil therearenow 149LNGships in active service,with another 27 on order. Thisr~uthberis likely to continue increasingas new LNGtradesarestarted.

The review of the marineincidentsfor LNG carriersinvolving the loss of containment during loading,transportationand dischargingconfirms a goodsafety record. The rate of seriouscasualtiesper shipyear for LNG carriers is approximately one half ofthat for other liquefied gas carriers. In addition thenatureof the incidentsinvolving LNG carriers wereof a minor nature comparedwith thosefor othervessels.

A review of historicalevidence(Ref. [11,[21) indicatesthat since1965, only a relatively small numberofspillage of LNG have occurred. In all cases thespillagevolumeswereminor. In total thesespillshaveresulted in:

• 2casesof severedeckfractures• 5 casesof minor deck/tankcover fractures• 2casesof lnvar membranerupture.

Four of theincidentswere dueto valveleakage.Suchincidentshave resultedin improved valvedesign.Inthe incidents thathave beenreported, therehasbeenno loss of life, damage to land-based property orharm to the environmentandon each occasiontheLNG dispersedwithout igniting.

There are no recorded incidents of collision,grounding,fire, explosionor hull failure which haveresultedin cargo spillage andno LNG carrier hasbeen lostat sea.

Evenin the two groundingincidentswhere bottomdamagehas occurred, therewasno releaseof LNG.

Designfactorshave minimised the consequencesoftheseincidents.Theseinclude:

• double hull protection• containmentsystemsspecificallydesignedfor

the transportationof LNG at low temperatures• transportationat atmosphericpressure

The above safety record demonstrates thatmaintaining safety is a principal aim of the LNGmarinetransportationindustiy.The 1MO GasCarrierCode and LR’s Rules for Gas Ships [3] providerequirements for design, construction and theequipment these vessels should carry so as to

minimise the risk to the ship, its crew and theenvironmentwith regard to thehazardsnvobreci

Thus LNG hasan extremely safe record in terms ofaccidental release. When the very few occasionswhen accidental release has occurred, theconsequenceshave beenminor.

RELEASE CONSEQUENCES- LNGEXPERIMENTS

Much of theexperimental work associatedwith LNGwascarried out in the1970’sand1980’s. Experimentswereundertakento gainabetterurtderstandingofthebehaviour of cold densegaseswhen releasedfromcontainment A further ol~ectivewas to study thecombustion characteristicsof LNG vapour. Theresults from the experiments were u.~edin thevalidation and development of computationalmethods for predicting the behaviour of thesesubstances.Therewas also a- need~to-confirm thefeasibilityof jettisoningcargosafelyif required.

The testsconcentrated on vapour cloud dispersion,vapour cloud ignition, pooi fires and rapid phasetransitioni.e. the instantaneouschangefrom liquid tovapour.

Dispersion

Note that the following dispersion theory dealspurely with gascloud movement. It doesnot takeinto accountthe ignition of a gascloud prior to itreachingits full dispersionrange.

Dispersiontrials on water (Maplin Sands, ThorneyIsland, China Lake and Burro/Coyote)(4M51,[61,(7M8M9J) showthatan LNG releaseresultsin a low lying heavycloud of vaporisedLNG withwell defined edgesmade visible by condensedwatervapour.This gascloud canbe used to indicatewherethe dangerous Lower Flammability Limit (UL) isfollowing a releaseof gas.

The LFL is the smallestamountof gasairratio thatcansupportcombustion(5% for LNG vapour). Henceit is importantas it indicates theareawhereignitionof the gas cloud could occur resulting In a fire.Generally when analysinggascloud dispersion, thelimit of the danger zoneis takento be half theLFL(i.e. 2.5% of volume). Both the 5% LFL andthe halfLFL occurwithin the areaof the visible condensedwatercloud. Hencetheignitabledangerousareaof agasplume is visibleatgroundlevel.

CONFIDENTIAL

I hi ctatinutil I ‘. opt nit in’, dod t toil ui i Ins .u i~ I lit isv ol I he iii thi Ii 1nil ni it itt I his ii s Regi‘ hr i t( Shipping. Lkivds

Register01 ‘iii ppi ng is not respitiisihle u 1 iible tor tiiv ii uhi U r,n. i irri r lit nets r eausid, relating to thesubstanceottheartide An~iii ii n~ on ~i iii thingin th is d rti lv isat thesi ~k i i skot tin. persi in who reliesi m it —

$

3

Experimentshaveshownthat the LFL for vaporisedLNG for a 20m3 releaseover 10 minutesis typicallybetweenIlOnt and150m (propane ranged between210mand400m). Thesedistancesincreaseup to 4ClOxnfor a40m3spill.

Largerspills of up to 20Om~overa similarperiod froma shipboardjettison systemalso resulted in inferreddownwind distancesto LFL of up to 400malthough avisible doud extending up to 2000m is possible lneach test the cloud height- was found to be in theorder of 10 to 12m.

Larger releasesproduce largergascloudsandhencelonger releasedistances. Distancesof around6kmhave been suggestedas a potential range from a25,000m3 spill. This figure does vary widelydependingonthe calculationmethodused -

B Bifurcation of a gascloud can produce fingers ofhigher concentrationgas than the averagepredicted

for thatdistance.

All vapourdispersiontestscarried out on fiat groundand water surfaces- are acknowledged to giveconservative results. The effect of obstructions,barriers, etc., would have the effectof reducing thespread of the cloud due to improved mixing andhigheratmospheric turbulencelevels.

Ignition

Ignition trials on dispersed unconfinedLNG vapourclouds have confirmed that no significantoverpressuresare developed[4J,[5J. Flamespeedsarein the order of 1Gm/secandmeasuredoverpressureslessthanImbarandtheflame maynot be sustainedthroughout thewholecloud.

in order to produce high flame speeds(i.e. >

lOOm/sec)in a methanegas cloud a high degreeofconfinement andcongestionis required, f~re~xamplein andaroundbuildings, processplant and pipework.Overpressuresin this eventwould be damagingbutrestrictedto within theconfined region, dying awayrapidly in the unconfined part of the doud. To datethere have been rio reports of a detonation in anunconfinedmethane gas explosion. Flame fronts inunconfined LNG vapour clouds have been observedto extinguish andnot propagate throughthe wholecloud, or stopandbe held stationaryby the wind atsome point away from the source,In seven LNGcloud ignition teststhe flameburnt back to thesourceon only one occasion.

$

PoolFires

Fire tests using poo1s of LNG up to 35jn diameterhave been carried out tiOl and measurementsofemissivepower show thatabove2Gm poo1

diameterthe emissive power reaches a maximum ofapproximately250kW/z&. It shouldbenoted that thevalue of emissive power obtained from testmeasurementsis dependanton an idealisedflameshapewhich must be adopted for the calculations. -

Values of emissivepower from different testsshouldonly becomparedif the sameidealisedflameshapeisadopted. -

For very largefiresthegenerationof smokelimits theamountof radiated heat. This can result in muchlower emissive heats in the range of around

-50kW-/-rn2. - - ---. -

250kW/m2

is the maximum emissivepower from theflame itself. This is an extremely high value(approximately 1000°C)when compared to otherchemicals. To put this into context, a humanimmediately next to such a flame would be killedinstantly. However thevalue drops the furtherawaypeopleare. 50kW/rn2 causesfatalityafter10 seconds, -

20kW/rn2 is enoughto causepainonexposedhumanskinafter2 seconds.1.5kW/rn2 is consideredsafe. Fora small fire (25m diameter) this safedistancewouldbearound 25Gmawayfromtheedgeof theflame.

Rapid PhaseTransitions

Rapid PhaseTransition (RFF) occurswhen LNG thathas agedin storage,due to relief ventingof vapour, isreleasedonto water. Alternatively, if a volume ofLNG (O.5rn3 andabove) is releasedonto water it agesdue to evaporation andcan undergo a RPT after adelay of severalseconds No igniticn is associatedwith the RFI’ effectandit has a limited capability fordamageto structures due to the physical explosiveeffects[9J. -

Multiple RPTsof varyingstrengthscanoccurover thearea of the release, the shock waves from eachcontributing to the- initiation of others. Damagingoverpressuresoccuronly verycloseto thesource.Noignition of vapour has beenobservedduring an RPT.However, ignition of the gascloud as a result of RPTdamage to neighbouring equipment orinstrumentationhasoccurred [11].

CONFIDENTIAL

— thest,itenuents,opinions niud etmclu~jtins it, those ‘it the tuth, ii andni it itt I l~iv~tsltgi.ter ut SIi~ppiiig. I_Iovt~

Registerol Shipping is not responsibleor li.tbh& or tue ti,iitttitu\ ‘ut error, Itt 0515 yr ~ttjsed,relating It’ thesubstanceut

thearticle. Any reliancetin ins thiitg in 11th. ti’thlt is ~it Ilti’ stile cisk it the~ ‘rsttn svlio reliestin it —

Releases onto water around jetties of a piledconstructionmay involve the flow of the LNG poolandvapour cloud beneaththe jetty. The vapour inthis casecould engulf the construction and anyshelteringeffects will prolong the duration of thecloizd. In the eventof anRPTbeneathajetty, damagecould occur. However,a potentialdanger resultingfrom RPTs is crackingof the ship’shull as aresultoflow temperature embrittlernent due to contact withtheLNC.

Experimental Resultsin Context

The experimentalwork conductedthroughout thelast25 years has resulted in a comprehensiveunderstandingof the hazardsassociatedwith LNGandtheconsequencesof bothignited andnon-ignitedreleases.This understandinghas therefore led togreater confidencein the accuracyof the modelsdevelopedto analysethe effectsof LNG releases.Theresults of the testshave also hadan impact on thedevelopment of operational ~and emergencyprocedures associated with the transfer andtransportationof LNG.

These tests must be placed into context with thesituationafter a releaseof LNG. If the ship is at seathen there is unlikely to be any local ignition sites.However there are also no general public orequipment in the local vicinity which could bedamaged. -

If the releaseoccursdoseto shore thenthere is likelyto be manysourcesof ignition oncetheLNG vapourcloud reaches land. These sourcesof ignition arehighly likely to causethe gascloud to ignite before ittravelsany significant distance. Thusthe damagewilllikely belimitedonly to theimmediateshoreareaandnofurther beyond.

Although such combustion may not burn back tosource,it is likely thatthe sheersizeof thespill, alongwith the numberof local ignition sourceswill causemultiple ignitionsof thegascloud. It is expectedthatat least one ignition will burn back to the sourcecausingapool fire to develop.

This review hai drawn from a numberof sources,both historicalevidence,experimentsandtheoreticalmodelling. This has resulted in the followingcondusions:’~

CONCLUSIONS

Historically for all typesof Lt’TG shipping therehasbeen noreported incidentsof lossof life onboardtheLNG ship, damageto landbasedproperty,or damageto the environment Designandoperatingstandardsonboard LNG ships have allowed only a smallnumberof releasesto occur. In each ca~ethere hasbeenminimal damageto theship.

Ignition and sustained combustion of a vaporisedLNG cloud under normal release conditions isdifficult. However the given a large number ofignition sources the gas doud -will probably igniteandeventually burnback to the liquid pooi it wasvaporising from. This will causethe pool to ignite.

Unconfined LNG vapour cloud detonation typeexplosionhas notbeendemonstratedin experimentalwork and is mostunlikely in practice.

Confined explosions could result in overpressures,-~but theseeffectsarelimited to the confinedspace,and:the effectswould dissipateawayfromtheevent.

The LFL for methaneair mixtures is 5% volumeso theLFL boundaryis well within the visible cloud atgroundlevel.

If a gas cloud is formed, and assuming that n~ignition occurs, the flammable limits have beensuggestedasreaching a substantialdistancefrom thesource. Such a value is only valid for a ship at sea.It is extremely unlikely that thesedistanceswouldbe reachedwhilst closeto shore or at berth due tolocalignition sourcesbeingreadilyavailable.

----—5

As the gas cloud warms up the gas will becomelighterthanairandwill riseawayfrom thesurface.

As LNG will vaporiseand is non-toxtc there is nosignificantdirectenvironmentaldamagecausedby aspill andhenceno deanup costs other than thosearisingfromsecondaryescalationfactors.

Thus it can be concludedthat LNG has specificparameterswhich make the likelihood of a majorexplosionremote. Ignitionsourcesproportionalto thesensitivity of the locationmeanthat gasplumesareextremely unlikely to pass long distances throughcities beforeigniting. As the gascloud warmsup itwill riseawayfrom thesurfaceuntil it dissipatesintotheatmosphere.

CONFIDENTIAL

- thesItteuttents,uuj’iniuns, uuutt u’tielto.i, ins .in Ilt,tse tul the.iutltttr and nut ot lloyd’s Registerti( Shipping. I_kyd’s

Ru’uister ot Stutppittgus not rt’spttnsiltk’ ‘r lttNy iii tue jiu,s to ri s ‘r ,‘rr,ur hut isv,’s ere,iused,relating U, thesubstancei ut

thear!i,le. Ane ut’hiuttce in Otis tltiiug iii lIds trtkIs’ is ~ui Itt,’ ‘l~’ risk ot thepersonwho relicson it —

4

I

a

5

Thesefeaturescombined with the high standardsofdesign nd operationthroughoutthe industry meanthat comparedto other chemicalsLNG posesoneofth~lowestthreats to thegeneralpublic andproperty.

REFERENCES

[1] INTERNATIONAL CHAMBER OF SHIPPING(1978). Tanker Safety Guide (Liquefied Gas).London.ISBN 0906270014.

[2] 6 SOCIETY OF INTERNATIONAL GASTANKER & TERMINAL OPERATORS(SICTFO).GasShippingIncidentReport (1999)

[3) IMO GasCarrier Codeand LR’s Rules for GasShips: Rules and Regulations for theConstructionandClassificationof Ships for the

Carriage

of Uquefied Gasesin Bulk. Uoyd’sRegister of Shipping,London.July 1986.

[4] JENKINS, D.R., MARTIN, J.A. (1982).RefrigeratedLPGsSafetyResearch.Gastech8Z

Paris,France.5-8November1982.

[5] BLACKMORE, D.R. etal. (1982).Dispersion and- Combustion Behaviourof Gas CloudsResulting

from Large LNG andLPG spills on to the Sea.Trans.LMar.E. Vol 941982Paper29.

[6] HEALTH AND SAFETY EXECUTIVE (1984).HeavyGas Dispersion Trialsat Thomey Island-

SummaryPapers.Sheffield, UnitedKingdom. 3-5April, 198t

[7] SCFJNEIDE1~A.L et aL (1979). US CoastGuardLiquefied NaturalGas Researchat China Lake.Gastech78. Monte Carlo, France. 5-8 October1978.

[8] PARNAROUSKIS, M.C et aL (1980). Vapour

Cloud

ExplosionStudy. LNG 6. Kyoto, Japan. 7-10 April 1980.

[9] McRAE, T.G. (1983). Large-ScaleRapid PhaseTransition Explosions. Lawrence Livermore

National Laboratory, Livermore, CA, UCRL-88688.May1983.

[10] N]—DELKA et al. (1989). The Montoir 35mDiameterLNG Pool Fire Experiments. LNG 9.Nice,France.17-20October 1989.

[11] ENGER, T., HARTMAN, D. (1972). Rapid PhaseTransitionDuringLNG Spillageon Water. LNG3.Washingtort~USA.24-28September1972.

CONFIDENTIAL

- This’ st,tteuiuents, ,t1

liititins, and ctuti~htisi~tnstn lhuvisi it this- ,iu(hui ir and iii ot lbuedsRegisteroh Shtipptttg. I.lt teds

Registeruul Shippingus not us~spon.subhs’sir Ik-ubh’ t ir ,utu\- ii a, tr,t. ou u’rr,r, bit itvu’\ er,ot.ust~st,relating lii Itt subslaneeitttheartieR’. Any reliancesin any thing in 1ltI~art is Ic is at the stile rtsk itt this: perstin who rehie~tin it —

6

-I - -

S

CONFIDENTIAL

- - I hi sb,itiiuis’iuls, itpttutons. ,uutd sttncbitsisuts it-,’ thui’se it Itt’ ,tttthtsir ,un,l not itt Ibm d’s Registermit ~hippuiig.hinyds

Registerat ~-mbuupptuugi~tint rs’spm mitsulibs’ i ir hu,uhube or ,ti us ituat mu ra~vi ir s’ritur, he ‘wi’s s’r caused,i-elating Ii’ thesubstancemit hut it Itt hi \sts ii It ttts i m ii tin thiutt iii thus ii bus Ii us it bht tilt uisl~mit th~P ssmn us Ito nut’, unit