Thispaper was presented at the 12th Annual OTC in Houston,Tex., May5·8, 1980. The materialIs subject to correction by the author. Permission to copy Is restricted to an abstract of not more than 300words.
ABSTRACT
Developments in f lare systems for the
emergency de-pr es su ri sa ti on o f oil productiont rains on offshore platforms in the North Sea arereviewed in terms of safety , efficiency, cost ,installation and material usage. The most commonlyused syst ems ar e enumerated and th e fundamental
requirements re-stated. Appraisal is made and
notional costs given of each type cited.
API guide RP521 rules for the calculation ofradiation and temperature levels ar e examined and
th e extrapolations to them r eq uired for use onNorth Sea platforms stated.
Structura1 prob1ems in the design oflightweight structures subject to ic e and high windloadings ar e examined with regard toa) dynamic analysis of individual members.
b) c on si de ra ti on o f wake induced forces andshielding.c) the derivation of wind force spectrum fromconsiderations of wind structures fo r inclusion ina dynamic analysis of complete structurese) selection of heat resistant materials inlocations near th e heat source
INTRODUCTION
The exploitation of o il fields in th e NorthSea has led to si tuations which th e generallyadequate design guidance given in th e API guide forpressure rel ief and depressuring systems (RP521)does not cover specifically. I ts intent has
therefore been ex trapo lated in various ways and th eobject of this paper is to review and to make a
personal appraisal of what has been done to datewit h r eg ar d to one aspect, namely th e positioningof burners and th e concomitant structure whichsuppo rt s t he ga s l ines to t he burne rs .
single platforms. Where p ipel in es t o shore were no t
possible the se faci l i t ies had to include st orage i n
addition to drilling, primary separation andproduction.
2) On single pla tforms the separation of hazardous
areas from accommodation and communication areascould no longer be acheived by separate platforms
inter-connected with bridges bu t had to be by fire
barriers dividing a single platform. This
necessitated t ha t e it he r th e gases to be burnt weretaken directly from t he p roce ss area and flared fromnear that hazardous area on th e platform o r the gaseswere conducted away by sub-sea l ine to a stackmounted on a secondary p1atform or swive11ing buoyremote from th e main structure.
For shallow, calm water th e remote s ta ck sol ut io n
combines safety with economy. For deep, rough watermaintenance problems concerned with purging
accumulations of condensate in t he sub -sea l ine andigniters and pilots sharply reduce safety standards,while th e placing of th e l ines and remote stacks indepths of water up to 150 metres multiplies th eminimum cost by a fa cto r o f at least 8.
A further complicating factor is that , due to thepresent climate of energy conservation, large volumesof ga s are only flared in an emergencydepressurisation or during commissioning of although th e pressure re l ie f system must assuredly be
available fo r an emergency, i t is otherwise l ightlyused. A system in frequent heavy use must be kept inef f ic ien t working order, one l ightly used is l iable
to receive less attention.I t is
therefore importantto keep depressurising systems free of diff icul t bu troutine maintenance tasks.
DESCRIPTION OF SYSTEMS IN USE
The principal types of f lare systems which have
been developed fo r North Sea use ar e shown in Figures1 to 6.
The main special circumstances in th e North
Sea which have influenced faci l i t ies design are asfollows:1) In deep, storm prone, waters economicsdictated that complete faci l i t ies be provided on
Fig. 1 Dual booms
References and i l lustrations a t end of paper.
557
The most compact and economical solutionconsisting to two short (up to 75m (246 ft) booms
placed to point outwards from prevailing winddirection, which should bisect th e angle between
th e booms. When th e wind shif ts to a point along
one boom and towards the platform ( i .e. directingthe flame towards th e platform) th e other boom and
burner carr ies th e whole flame load. The only
disadvantage of this system is in th e danger ofmis-d irect ion of th e burner to be used in adverse
wind conditions. The elevation of the booms to ahigh i nc li na ti on w i ll allow th e flame path toclear th e platform and derrick but permi ss ib le
r adia tion leve l s could be exceeded on the workingareas. Radiat ion screens can reduce the level ofradiation in affected areas such as th e upper·
le ve ls of th e dr i l l derrick. A great advantage ofthis system is that f lare t ips can be changed andmaintained without interfering with production.
metres (1000 f t ) . Problems must arise with keeping
th e l ine free from condensate and in maintenance ofpilots , igniters, thermo-couples and flaret ips.
Table 1 shows th e approximate ranges of materialweights and f abr ica tion cos ts fo r each of th e sixoptions, assuming th e average water depth of 140metres (460 f t ) , environmental condit ions applicable
in the North Sea area, and a maximum f lare capacityof 460,000 kg/hr (1,000,000 lbs/hr).
High p i tch cant il eve r s tend to be cheaper thanlow pitch as t hei r e f fec ti ve length in a horizontal
direction is less, moments due to dead and ic e loads
ar e reduced which allows fo r reduced section' sizeswhich in turn reduce wind loads,
Fig. 2 vert ical stack
Vertical stacks from th e platform r equi re t o be a t
least 100 metres in height. They general ly requi reto be instal led in pieces offshore and their bases
take up a lo t of valuable pla tform space. Acruciform tower offset from one corner of th e
platform could be instal led in one piece and wouldnot reduce th e usable p lan a rea of t he p la tf orm.
Maintenance facil i t ies are superior to thosepossible on outward cantilevering stacks and flare-t ip changing could be a rr anged wit hout acces s to
the to p platform of the stack. Thus, with twinf la re li ne s, t ip s could be changed without inter-fering with production.
Radiation levels can be kept to acceptable l imitsregardless of wind direction. I f l iquids are not
completely eliminated from th e vapour t ra in there
is a small possib il it y that in l ight winds, burning
l iquid part ic les may reach the platform deck.
Bridges require to be a t leas t 182 m (600 ft) long
and to have a stack a t the end a t least 33m (100 ft)above top deck level. The supporting tower fromthe sea bed might need to be as long or longer thanth e bridge, depending on water depth, and must bedesigned to carry maximum wave loads. The systemhas the advantage of being able to accommodateemergency generation plant and equipment remote
from th e main platform.
To be effective such single booms require to be a t
leas t 100 metres (328 ft) in length. Stayedcanti levers are the mos t economica l structuralform. Flame shapes relat ive to these two types areshown in Figures 7 and 8 discussed la ter in th epaper. For th e high pi tch canti lever , r adia tion toth e working areas can be kept to an acceptablelevel and there is l i t t le danger of l iquidpart ic les being carried back on to t he p la tf orm.
The boom may be instal led in one piece but th eoperation is by no means a s imple one , and is besteffected from th e stable p la tf orm off er ed by asemi-submersible crane vessel.
The length of th e l ine is generally about 300
Fig.
3 & 4
Fig 5
Fig 6
Low and high pitch cantilever booms
Bridge on secondary support
Sub-sea l ine to a remote stack
558
Only fabrication costs are given asinstal la t ion costs are generally contracted for thecomplete top-side and i t is therefore diff icul t to
extract f ig ur es f or any single item of equipment.
A notional assessment of th e instal la t ion costsincluding transportation would be :
1) Options shown in Figures 1 to 4, 0.5 to 0. 8
times th e f ab ri ca ted cost . The causes of variation
would be
a) Downtime due to·weather in a l l cases
b) In the case of th e vert ical stack, th e size ofindiv idua l p ieces of tower to be l i f ted.
2) Option shown on Figure 5 between 7,000,000 and9,000,000 dollars.
3) Option shown on Figure 6 is complicated by the
add it ion o f a short length of subsea l ine requiringa t least two heavy anchorages, between 12,000,000 and14,000,000 dollars.
APPRAISAL OF SYSTEMS
If a sat isfactory degree o f p ro te cti on fromradiation and heat effects is assumed fo r a l l types,neither of the options shown in Figures 5 and 6appear very attractive economically, their instal ledcost bei ng abo ut te n times that of th e others andmaintenance (particularly of that with a sub-sea
line) would show a t least an equal level ofescalation.
Instal led costs of th e other types are moreclosely bunched and must be considered in context ofth e other parameters making a value jUdgement.
The sole disadvantage of th e dual boom system is
in th e possibil i ty of an emergency rel ief of gas
being directed over the platform when th e winddirection is parallel to one boom and t he pla t form isleeward to that boom. As one of th e objects of thissystem is to allow th e tota l f la re in g t o take placefrom one selected boom i f pressure rel ief must betaken under adverse wind conditions, i t may appear aninconsiderable risk th at such a sys tem should fai l .Nevertheless there is a strong body of opinion amongprocess engineers that an emergency pressure rel ief
system should work infal l ibly without manualselection.
I t is apparent that flame from an hor izont al o rlow pitched boom could actually impinge on the
pla tform or a t least on th e dr i l l d er ri ck i n th e case
proposed. I f the booms ar e elevated to a 45° angleand spring from th e highest level. possibJ .e on th etopside the fJ.are t ip will invariably be abovederrick head, th e working levels of which can be
protected by radiation screening. This arrangement
would provide a fai l safe device in th e event of am is-application of the flareboom selectionprocess.
For t he p rocess engineers whose objections toa manual selection system is too strong toovercome, th e remaining systems shown on Figures 2,
3 and 4 have been u sed, w ith th e exception of th ecruciform vert ical stack. Their cost installed areof th e same order.
One disadvantage of a vert ical stack is thati t is liab le to take up 8 to 10% of th e availabletop-deck space and cannot generally
a) be placed in i ts most logical position over
t he well -head area as i t would interfere withth e dril l-derricks
b) be placed between th e well-head and th e accom-modation areas without reducing. crane
coverage over the deck in addition to moving apotent ial hazard closer to accommodationareas than desirable.
A cruciform vert ical stack se t off one corner
of the p la tform overcomes this problem.
The low p it ch s ingl e c an ti le ve r boom suffersth e same ( if diminished by virtue of i t s greaterle ng th ) d is advanta ge a s th e low pi tched dua l boom.However, i f an adverse wind is blowing stronglyfrom t ip to platform there remains no al ternat iveflare path and, depending on the length of the booma hazard or a t leas t discomfort from smoke cloudcould exist.
The high pitched single cantilever can
overcome a l l th e radiation problems to th e deck and
dr i l l r ig, bu t requires a platform shut-down to
c arr y out a t ip changing. or maintenance operationand is certainly more diff icul t to design,fabricate and instal than any of the o th er v iabl eoptions. Each type ful f i l s neatly a l l of the idealrequirements, consequently selection is a matter oflogica l in terpreta tion of part icular cases. In th e
author's opinion the basic cr i te r ia fo r th eposi t ion ing of a flare burner should be as f ol lows :
i ) The burner should be placed a t an elevation toavoid a flame impingeing on the dr i l l derrickwhen th e flame is streaming (Le.horizontal) • No provision is made againstthis case in the guidance afforded by RP521but i t can occur i n p rac ti ce .
i i) The burne r shou ld be be leeward from the
platform for th e most prevalent winddirection.
i i i ) Tip changing or maintenance should,preferably, be possible without necessitating
a platform shut down.
iv ) Radiation levels calculable according toRP521 should be observed with the provisionthat fo r cases of remote probability (e.g. aful l emergency de-pressurisation with th e
559
most adverse wind possible) acceptabler ad ia ti on l evel s could exceed those given in
guide by a factor of up to two.
These cr i te r ia assume that emission of liquidpart ic les is virtually eliminated by an eff icientknock-out system.
GUIDE ROLES RELATED TO BURNER POSITION
Only matters related to burner position and
supporting structure are dealt with.
Guidance is given fo r the case of vert icalburners only and radiation is calculated from a heatsource a t the mid-point of the flame.
The calculation method is specifically intended
to provide the minimum height of vert ical stackrequired from grade. In th e case of an offshoreplatform i t is often considered convenient to i nc li neth e stack and burner a t an angle which may varybetween 0
0and vert ical . The dual effect of this is
that the nearest sensitive point on th e platform mayno longer be the equivalent of "grade" and the mid-
point of the flame may no longer be an adequat e
representation of the heat source (which i t is in th ecase of a vert ical stack).
I t should be noted that a vert ical burner isgenerally accepted to be best in terms of long l i fe
and maintenance.
Another and almost equally important effect froman inclined t ip is that the flame shape is influenced
by th e gas velocity at th e orif ice. Where flaring isat a low rate (of purge gas only, fo r example) and ahigh wind is blowing back a long th e bu rn er l ine thereis a high probabil ity of :i) Flame lick back along the burner reducing th e
th e burner l ife
ii ) The flame "streaming", Le . taking an horizontalpath instead of the rising curve derived from a
combination of a gas and wind velocity andconvection effects, but as an inclined burner is
usually positioned a t a low elevation th ephenomena can be more cr i t i ca l in i ts case.
One advantage which occurs in platforms remote
from th e shore is that wind turbulence a t th e levelsa t which burners are placed is generally much lessthan on-shore, where topography can induce high threedimensional turbulence. Provided th e tip hassufficient elevation there is only a low probabilityof th e flame being swept below t ip level.
The theoret ical flame shape from an inclinedburner may be easily computed from an extrapolationof the formula given in th e guide. I t would appear
prudent to conside r t he effects of a streaming flamefo r cases in which th e gas velocity is low comparedwith the wind velocity.
Figures 7 and 8 show flame shapes derived by th eRP52l formula fo r th e conditions relevant to th eflare booms shown in Figures 3 and 4 and the
previously quoted flare capacity. Envelopes of flameshape are shown fo r 20% to 100% flaring capacityunder adverse winds of 10 and 70 mph wind speed. Forth e 25
0boom a t higher wind speed th e t a i l of th e
flame would point directly a t the upper le ve ls o f th e
dr i l l derrick. In the case of th e shorter dual boom
of low pitch the flame tail could impinge on thedrill derrick if the wrong boom was used byaccident. Both of these situationscouldbe lethalto people working on the derrick regardless ofradiationscreening,althoughit is justwithinthebounds of possibility that the radiation andtemperaturelevelas calculatedfrom theprescribedheat sourceat flame mid-pointwould not be wildlyhigh.
Another factor to be considered regardingflame shape from low pitched tips is that under
adverse conditionsthe flame passesmuch closer tothe flareline and its supportingstructurethan isthe case for a vertical stack. High radiationeffectsmay thereforebe expectedin the area nearthe end of the boom givin rise $0 temperaturesinthe bmm structureof 300 to 350 C (572to 662°F).Such temperatureswill sharplyreducethe allowablestresses in carbon steel and the life of anycorrosion protection coating. In view of thedifficultyof maintainingcoatingson booms and theprocess down time such maintenance entails, thelatterproblem is the more serious.
Calculationsfrom a single point heat sourcetend to show an optimistic temperature levelwhereas calculations from heat sources equally
distributedalong the flame tend to be pessimistic.The reason for pessimistic results from uniform
multiple sources is that the radiation from thenarrow pencil of partially burnt gases near theburner is considerablylower than for the greaterdiameter flame front farther back in the flame.That is, multipleheat sources from a single flame
shouldnot be treatedas uniform.
STRUCTURALDESIGN
The design procedure can be divided into two
parts:
a) The application of static wind and iceloadings in combination to establish the
structuraltopologyand sectionsto be used.The treatmentof forcesdue to these types ofloading is well established and needs nofurther comment. Maximum wind speeds in theNorth Sea are sufficientlywell documented,but estimates of ice accretion vary amongauthoritiesfrom a maximum of 100MM to zero.The only thing the estimateshave in common isthat the ice thicknesswill be greater at lowlevel and will diminishwith height. It mustalso be said that the location of thestructure in the North Sea has an importantbearingon the level of ice accretion.
b) Using the structurearrived at by the staticanalysis, a more refined analysis must be
carried out to check for dynamic and fatigueeffects.
Dynamic Analysis
A wide ranging documentationof the problemswhich may be encountered may be found inEngineering Sciences Data Unit series on WindEngineering,Volumes 1 to 4. The coverageextendsfrom wind turbulence data, mean loads onstructures, fluctuating loads, response and thenatural vibrationparametersof structures.
The actual analysis may be divided into tparts:
i) Verificationof the individualmembers in tstructureagainst susceptibilityto vortex sheddiresonance. A convenient starting point for thinvestigation is the graph shown on Figur9 which should be used for chords of booms andsimilar graph but for the fixed ended condition fbracingmembers.
These graphs compare the natural frequency otubulars of normal wall thickness to the Strouhfrequencywhich definesthe vortexsheddingfrequenfor specificwind speeds for each diameter.
From this graph one can readilycheckmembersangroups of members to assess if they are in thcritical range. Members in the critical rangwill not be subject to resonance provided thinequality
is satisfiedwhere
M = mass per unit length
P =air density
D = member diameter
and ~= = logarithmicdecrementof daming.
The last named factor for welded steel tubulstructures can vary between 0.001 and 0.01. Thlowestvalue of the dampingparameterwould applya member which is in a more or less steady state of
tension whilst being subjected to vortex shedding
from wind (e.g.a tie carrying a heavy dead loasubjectedto a cross wind). The highest value woul
apply to members subjectedto rapidly fluctuating
random stresses (e.g. bracing members in the maistructure). Welded structuresgenerallyhave lowdamping characteristicsthan do bolted or rivettconstruction. In practiceaxial forcesin membersdnot significantlyaffect the natural frequencyofmember, a differenceof ~ 5% may occur.
It is possible to quantify the steady staamplitude of a resonant member by a somewhlabrynthine tabular method contained in E.S.D.ItemNumber 78006 but only in an investigationof tcause of an actual case of resonance is suchcalculationappropriate.
Although lattice structuresare not generalprone to vortex sheddingresonanceas air flow ovthem tends to be broken up by the bracing memberwhere two dominant members of like diameter arparalleland placed less than six or seven diameteapart and contain a susceptiblenatural frequencyis possiblefor resonanceto be set up in one or botmembers,the frequencyof resonancemay be very higThe phenomena can occur with chords of a lattistructureor twin gas lines.
From this verificationany suspectmembers cabe adjustedby increaseof diameteror sub-bracingtensureno individualmembersare liableto resonancPractically this is the important part of th
analysis but if called for the second part may be The threecomponentsof load effectare takenasentered. those from
ii) Analysis of the whole structure can then be a) the one hour mean windundertakenusing as near to a member for membermodel as possibleeconomically,and extractingthe b) fluctuatingnon-resonantforceseigenvaluesand theirrelatedfrequenciesand modalshapes. c) fluctuatingresonantforcesfrom which theremay
Three main types of wind induced dynamicbe significantcontributionsfrom more than onestructuralvibrationmode.
effectscan relateto the main structure,the firstis due to vortex sheddingon the dominant member
(invariablythe flare lines), the other two being
For remoteoffshoreconditionswith long fetches
the surface roughness parame~~due to wind turbulence buffeting and wake
& ~o_z;d
buffeting.constant (generallytaken as 10 .
This reduces turbulence within the gradient
Assessmentof the vortex sheddingtype beginsheight and eliminates one variable from ‘the
with a comparison of the shedding frequenciesproblem,leaving the followingto be considered
i) Turbulence intensity and length scales ofpossible and the natural frequencies of the turbulencein the x,y and z directions,all ofstructure. which vary with height (of the structurefor the
specificcase))Rigorous treatment of wind turbulence
buffetingii) The natural frequenciesof the structure
requires modelling the actualconditions,initiallyit is enoughto considerthat
iii) The propertiesgiven in (i) in conjunctionwith
in the virtuallyundisturbedatmosphericwind foundthe range of eddy frequencieswith which the
at this height in remote offshore structuresthestructure is liable to react significantlycan
higher frequency eddies contain less energy thanbe used to derive a non dimensional spectral
those with a lower frequencyof occurrence.density function,a curve of which is shown inFigure 10. The summationof the valuesof thisfunctionfor the equivalentnatural frequencies
Treatment of wake buffeting requires
knowledgeof wind behaviourover the totalplatformof the structuredefinesthe resonantcomponent
iv]structurewhich can only be achievedby a physical
Mean mode generalisedaerodynamicorcesfor each
(windtunnel)model.mode to be consideredfor summationof the totalvariance. Each consistsof the summationof the
As the structureof a platformis liableto bemean forces on the boom times the normalised
sharp edged inmode shape.
main shape, its behaviour in v]windflow will not be subject to scale effect and
A structuraldampingparameterwhich can only be
ReynoldsNumber need not be matched between modeldefined simply for structureswith linear modeshapesand constantmass distribution.
and prototype.
OrI the only wind tunnel test of an aeroThe magnitude of these variables cannot be
elasticmodel of a boom attached to a platform ofassessed with great accuracy in the present state of
knowledge.which the authorhas knowledge,wake buffetingas a
It is probable that the mean forces canbe accurate to within ~ 25%. For most lattice
principal source of excitation was shown to be structures the dynamicsufficient to excite the boom in the lower
response contributions
frequency modes but the boom remained stablesummate to be nearly equal to those from the mean
throughout the wind speed ranges tested for allforces,and theycan bg subjectto the same degreeofdivergenceas can the mean forces. For the maximum
wind directions. and minimumcases this could lead to an error of plus
If the shedding or buffeting frequencies42% or minus 33% of the correct value.
appear to be close to any of the lower naturalfrequenciesa response analysismay be necessary.
It shouldbe noted that the sourceof error lies
Generallythe sheddingfrequenciesare too high andwith the actual wind speeds, turbulencescales and
the buffeting frequencies too low todampingparametersused thereforea static analysis
havesignificantresonant effects on a beom of normal
using 3 secondgusts would contain most of the same
stiffnessthough there will alwaysbe some dynamicerrors, but a static analysiswill not give accountof the reactionbetween wind turbulencefrequencies
amplification. and structuralfrequencies.
PROBABILITYBASED DYNANIC ANALYSIS Where the structure has adequate stiffness to
Much endeavour has been made to combine theresist periodic turbulencethe static and the prob-
researchcarried out on the nature of wind forcesabilistic dynamic methods of analysis give rise to
and the mass/flexibilityvery similarmaximum member forces.
characteristics ofstructureswhose principal external loadings aredue to wind forces. Because of.the number and
For flexiblestructuresthe dynamiceffects due
diversity of the variables involvedand the factto turbulencebecomemore marked and the value of the
that some of these have yet to be investigatedinprobabilistic method lies in checking that such
depth, the most logical approach appears to bestructures are not prone to turbulence buffetingresonance.
analysison a probabilitybasis for the fluctuatingFurther research on the value and
forces, which can then be applied as dynamictreatmentof dampingparameterswould constitutethe
responsefactorsto that most constantregisterofonly major refinementrequiredto the method, which
wind speed, the one hour mean.is in any case viable in its present form.
