... ,PB83-2637231111111111111111111111111111111111111EPA-600!2-83-070August1983EVALUATION OF THE EFFICIENCYOF INDUSTRIAL FLARES:Background - Experimental Design - FacilitybyDo Joseph, J. Lee, C. McKinnon,R. Payne, and J. PohlENERGY AND ENVIRONMENTAL RESEARCH CORPORATION18 MasonIrvine, California 92714EPAContract No. 68-02-3661EPAProject Officer: Bruce A. TichenorIndustrial Processes BranchIndustrial Environmental Research LaboratoryResearchTriangle Park, NorthCarolina 27711Preparedfor:U. S. ENVIRONMENTAL PROTECTION AGENCYOfficeof Research and DevelopmentWashington, DC 20460, - ~I REPRODUCED BY;~ I'" . u.s. Department or CommerceNational Technical Infonnalion ServicelSpringfield, Virginia 22161 ~_. -----./TECHNICALREPORT DATA(Plcau read bUJrlJctions onthe rercrsebefore completing)1. REPORTNO.12. 3.REPBSNiS AC2E6S372' eEPA- 600 / 2- 83-0704. TITLEAND SUBTITLE 5. REPORTDATEEvaluationoftheEfficiencyof Industrial Flares:August1983Background- -Experi'11entalDes ign- - Fac ility6. PERFORMINGORGANIZATIONCODE7. AUTHORIS) B. PERFORMINGORGANIZATIONREPORTNO.D. Joseph, J. Lee, C. ~ c K i n n o n , R. Payne, and J. Pohl9. PERFORMINGORGANIZATIONNAMEANDADDRESS 10. PRCIGRAMELEMENTNO.Energyand Environmental ResearchCorporation18Mason11. CONTRACT/GRANTNO.Irvine, California 9271468-02-366112..SPONSORINGAGENCYNAMEANDADDRESS 13. TYPEOFREPORTANDPERIODCOVEREDEPA, OfficeofResearch and DevelopmentPhase1. II: 10/80-1/82IndustrialEnvironmental ResearchLaboratory14. SPONSORINGAGENCYCODEResearchTrianglePark, NC 27711EPA/600/1315. SUPPLEMENTARYNOTESIERL-RTPproject officer is Bruce A. Tichenor, 'YIail DropOJ,919/541- 2547.~16. ABSTRACT The reportsummarizesthetechnical literatureontheuseof industrialflares andreviewsavailableemiss ion esti'11ates. Technicalcritiques of past flareeffic iency studiesareprovided. l\Il:athematical modelsof flamebehavior areexploredandrecommendations for flareflamemodelsaremade. Theparametersaffectingflareefficiency areevaluated, andadetailed experi'11ental test planis developed.Thedesignofaflaretestfacility isprovided, includingdetails ontheflaretips,fuel and steamsupplies, flowcontrol and measurement, em iss ions saTYlplingandanalys is', anddata acquis it ionandpFocess ing.-17. KEYWORDSANDDOCUMENTANALYSISa. DESCRIPTORS b.IDENTIFIERS/OPENENDEDTERMS C. COSATI Field/GroupPollution "'1easurement Pollut ionControl 13BExhaust Gases FlowControl Stat ionarySources 2lB 20DEfficiency Sampiing Industrial Flares 14G 14B1\1athematical Models Analyzing 12AFlamesFuels 21DHI. DISTRIBUTIONSTATEMENT 19. SECURITYCLASS(This Repo;/j 21. NO. OFPAGESUnclass Hied 2g"ReleasetoPublic20, SECURITYCLASS(Thispage) 22. PRICEUnclassified;" "-EPAForm22201 (973)i: 'IThis document has been reviewed inaccordance withu.s. Environmental ProtectionAgencypolicyandapproved for publication. Mentionof trade namesor commercial products does not constitute endorse-ment or recommendationfor use.ii,....ABSTRACTThe U.S. Environmental ProtectionAgency has contractedwith EnergyandEnvironmental Research Corporationtoconduct aresearch programwhichwillresult inthe quantificationof emissions fromand efficiencies of industrialflares. The studyis beingconducted in four phases:I - Experimental DesignII - Design of Test FacilitiesIII - Development of Test FacilitiesIV- Data CollectionandAnalysisThis report provides the results of Phases I and II of the study.The report summarizes the technical literature on the use of flares andreviews available emissionestimates. Technical critiques of past flareefficiencystudies are provided ..Mathematical models of flame behavior areexplored, and recommendations for flare flame models are made.The parameters affectingflare efficiencyareevaluatedandadetailedexperimental test plan is developed. The designof aflare test facilityisprovided, includingdetails on theflare tips, fuel and steamsupply, flowcontrol andmeasurement, emissions samplingandanalysis, anddata acquisi-tion and processing." AQualityAssurance/QualityControl Programis alsodescribed.Results of the testing program(Phases III and IV) will be included ina later report.iiiTABLE OF CONTENTS2.1 Use of Industrial FlaresSUMMARY .BACKGROUND2.4 Experimental Information on Flares.2-482-492-602-682-682-70-Page1-11-11-31-71-92-1 2-22-32-82-8 2-132-132-172-19 2-192-202-342-362-402-44 2-452-46PetroleumRefiningPetroleumProductionBlast FurnacesCoke OvensChemical Process Industry .Summary of Use of Industrial FlaresDesignof FlaresFuels Flared.. ."Flare Operating ConditionsFlare Sizeand Capacity ..Summaryof Commercial FlaresFlare Characteristics .Characteristics of Previous ExperimentalStudies on Flares .....The Structureof Flare Flames .Flare Efficiency .Productionof Soot in FlaresThe Effect of Wind on the Performance of FlaresThe Effect of SteamInjection/Forced Draft onthe Performance of Fl ares . . . . . . . . . . .2.1.12.1.22.1.32.1.42.1. 52.1.62.3.12.3.2 .2.3.32.3.42.3.52.4.12.4.22.4.32.4.42.4.52.4.62.4.71.1 Pollutant Emissions FromFlares1.2 Deficiencies in Previous Flare Emission Studies1.3 Technical Approach.1.4 Report Organization2.2 Emissions fromFlares2.3 Commercial Flares ..1.02.0SectionvPrecedingpageblank" ,-Section3.04.0TABLE OF CONTENTS (Continued)2.5 Modelingof Flares .....2.5.1 Models of Flare Behavior2.5.2 Previous Models of Jets.2.5.3 Solutions of the Transport Equations2.5.4 Scaling Considerations .2.5.5 The Broadwell Model of Turbulent Flames.2.5.6 Recommendations for Modelingof FlaresTHE NEED FOR WORK . . . . .TECHNICAL APPROACH AND EXPERIMENTAL PLAN4.1 Overall Approach .4.2 The Need for Studyof Pilot Scale Flares4.3 Size of Pilot Scale Flares4.4 Operating Conditions4.5 Selectionof Gases.4.6 The Effect of Steam4.7 The Effect of Wind.4.8 Experimental Measurements4.9 Modelingthe Emissionof Pollutants FromFlares4.10Experimental Plan .Page2-772-78 2-87 2-100 2-1032-1092-1133-1 4-14-14-44-5 4-10 4-12 4-12 4-13 4-134-154-164.10."1 RequiredScope of the Pilot ScaleTest FacilHy .4-164.10.2 Experimental Program . 4-17,-5.06.0FACILITIES REVIEW .....5.1 FacilityRequirements5.2 Existing Facilities5.3 Proposed Flare FacilityDESIGN OBJECTIVES .6.1 Design Criteria6.2 Approach ..6.3 Parameters ...viPrecedingpageblank5-1 5-15-65-7. . 6- 1 6-1 6-1 6-1TABLE OF CONTENTS (ContinuedSectionAPPLICATION AND ANALYSIS OF DATA7.8 Control, Data Acquisitionand Processing6.4 FacilityCapabilityFACILITY DESIGN .....8.1 Independent Variables8.2 ,Cal cul ati on of Emi ssi ons . . . .8.3 Interpretationof Flame Structure6-26-26-2. . 6-26-36-36-37-17-17-37-97-9. . 7-97-127-127-157-157-177-197-227-267-46. 7-527-537-588-18-1. 8-28-4, .Flare Size .....Flare Gas PropertiesNozzle Exit VelocityWind ConditionsAir EntrainmentMeasurementsFuel Metering .Tracer Meteri ngSteamMetering6.3. 16.3.26.3.36.3.46.3.56.3.67 ~ 5 . 17.5.27.5.37.1 Flare Stack and Flare Tip7.2 Fuel Supplyand Handling7.3 Tracer Supply .7.4 SteamSupply .7.5 Input FlowControls and Metering.7.6 Ambient Conditions Measurement and Control.7.7 Measurement Techniques .........7.7.1 Global (Overall) Combustion Efficiency7.7.2 Applicationof Tracer Gas7.7.3 Extractive Sampling .7.7.4 VelocityMeasurement .7.7.5 TemperatureMeasurements7.7.6 Characterizationof Flame Structures7.08.0vii".....Section9.0TABLE OF CONTENTS (Continued)8.4 Scaling and Modeling8.5 ConclusionsREFERENCESAPPENDICESAppendixA- Comparisonof Commercial FlaresAppendixB - Summaryof CARB (CaliforniaAirResources Board) Survey .AppendixC - QualityAssurance Plan .....Appendix D- Emission Factors for Flare CombusitonAppendix E - Calculationof Flame Shape and LengthviiiPage8-58-6 9-1 A-1 B-1 C-1, D- 1 E-1LIST OF FIGURESFigure2-1 (a) Variation in gas densityflared froma German refinery;(b) Actual floWrateof test flare used bySiegel (1980) . 2-62-2 Volume ratios of hydrocarbons in theflare and off-gas forthree tests on aflare at a German refinery (Siegel, 1980) .. 2-92-3 Components of an elevatedflare (Klett and Galeski, 1976). 2-212-4 Rectangularmulti-jet ground flare (Klett and Galeski,1976) . . . . . . . . . . . . . . . . . . . . . 2-.222-5 Designof flare tips (Klett and Galeski, 1976) . 2-262-6 Stack height and allowable radiation intensity (Oenbringand Sitterman, 1980) . . . . ... 2-302-7 Commercial flare heads. (a) 20 Mscfd pipe flame; (b) 60Mscfd smokeless flare anda 20Mscfd smokyflare,(Peabody/Kaldair, 1979) 2-412-8 Distributionof flare nozzle sizes reported by Californiarefineries toCalifornia Air Resources Board Survey (1980). 2-432-9 Estimates of flare emissions due to incomplete combustionof eddies . . . . . . . . . . . . . . . . . 2-522-10 Temperature profiles inacommercial flare . 2-532-112-12Effect of propaneemissions on combustionefficiency.Propane as fuel . . . . . . . . . . . . . . . . .Effect of CO emissions on combustionefficiency.COas fuel .2-572-58'.-2-13 The effect of throughput on concentrationprofiles intwo flare flames (Siegel, 1980) ' . . . . . 2-592-14 Radial concentrationprofiles in aflare flame (Siegel,1980) . .. . . . . . . . . . . . . . . . . ... . . . . 2-612-15 Species centerline concentrations as afunction of heightabove burners (Lee andWhipple, 1981) . . . . . . . ... . 2-662-16 Summaryof flare emission, excludingsoot, as afunction ofheight above burner tip (Lee and Whipple, 1980). . . ... 2-672-17 Effect of soot concentration on combustionefficiency.Propane as fuel . . . . .. 2-69ixFigure2-18LISTOF FIGURES (Continued)PageTheeffect of steaminjection on flame length (Siegel, 1980) 2-72.2-19 The effect of steaminjectionon concentrationprofiles,3and 6 meters above flame tip (Siegel, 1980) . . . . . 2-742-20 The effect of steaminjectionon local flare efficiency(Siegel, 1980) . . . . . . . . . . . . . . . . . . 2-752-21 Theeffect of steaminjection on temperature (Siegel, 1980) 2-762-22 Short time photographs.of turbulent flame ( B e ~ k e r , et al,1981) . . . . . . . . . . . . . 2-792-23 Conceptionof aflare shedding eddies . . 2-802-24 Reaction pnd life of eddies shed fromaflare . 2-812-25 Eddy frequency . 2-83:;-2-26 Concentrationof CO in eddies . 2-842-27 Decayof an eddyfromapool {ire (Bratz, et al, 1980) 2-852-28 Geometric dimensions of the largest and smallest eddies inn-hexane pool flames as,a function of pool diameter (Bratz,et al, 1980) . . . . . . . . . . . . . . . . . . . . .. . 2-862-29 Progressive change in flame type with increase innozzlevelocity (Hottel and Hawthorne, 1949) . . . . . . . . 2-882-30 Theoretical predictionof entrainment in buoyant jets(RicouandSpalding, 1961) . . . . 2-902-31 Entrainment by buoyant jets and flames (Ricou and Spalding,1961) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 912-32 Definitions of x, z, the curvilinear coordinate ~ , horizon-tal and vertical velocitycomponents uand w, velocity U,and flame radius 0 .................... 2-922-33 Computed profiles of compositionand densityfor d =0.005m,w= 22.1 m/sec, and uQ) - 2.55m/sec . . . . . . . . ... 2-962-34 Experimental soot concentrations on the axis of the C2H2diffus;'on flame (Re =7000) comparedwith predictions(Magnussen, 1980) 2-99xLIST OF FIGURES (Continued)Figure2-35, .Mixingfactor (M2) anddegree of oxidation (N) for"Constant Residence Time" scaled natural gas flames--axial traverses (Salvi and Payne, 1980) ..... 2-1082-36 Comparison of model predictions of NOmass fractionagainstmeasurements of Bilger and Beck (1975) for hydrogen-airflames for x/d > 30. The prediction indicates afrozenNO flux, whereas the data indicate NO destruction. . 2-1124-1 Experimental flare tipconcept 4-117-1 Six-inchflare head. . . . . 7-47-2 Gas consumption for different nozzle sizes and nozzles gasvelocities . . . . 7-67-3 Fuel supplysystem 7-87-47-57-67-77-8.7-97-10a7-10b7-11Flare - tracer supplyandmeteringFuel control andmetering systemSteammeteringWind generator, support structures, and sample hoodsExhaust gas collection hood for pilot scale flare ..Distributionof intermittencyfactor, n, and mean forwardvelocity inaround free jet (Corrsinand Kistler, 1954)Approximate streamlines ina turbulent roundjet. Thespreadingangle is shown for two values of intermittencyfactor, n .Airflowstreamline drawn intoa hood, fromAir PollutionEngineeri ng Manual, 1973 . . . . . . . . . . . . . . . .Errors in solid concentrationfor samples drawnanisokinetically (Badzioch, 1959and 1960)7-11 7-14 7-167-187-217-237-247-24 7-297-12 Water-cooledsoot-samplingprobewithexchangeablefiltersandfiltertips (Chedaille and Braud, 1972) .. 7-317-13 Sample systemschematic ' 7-37: ,.....7-14 Tedlar bag sample concentrations changes with time(Polasek andBullin, 1978) .xi7-38LIST OF FIGURES (Continued)Fi gure7-15 Design of multi-rake probe .. 7-407-16 Photographicanalysis. 7-57A-l Principal elements of a "steam-ring" smokeless flare tip 1976) .. A-2A-2 Principal elements of the Flaregas FS antipol1utant flaretip1976) . . . . . . . . . . . . . . . .. A-3'.--A-3A-4A-5A-6Principal elements of the Indair smokeless flare 1976) .Principal elements of the Smoke-Ban model SVL flaretip1976) .Principal elements of the Zink Services SAfield flaretip1976) ...........Principal elements of atypical groundflare (Brzustowski 1976) . . . . . . . . . . . . . . . . . . . . . .xiiA-4A-5A-6A-7LISTOF TABLESTable1-1 Previous flare emission studies . . . . . . . . . 1-4i2-1 Surveyof Californiaoil refineryflares (CARB, 1980) ... 2-42-2 Gas flared in U.S. refineries (Klett and Galeski, 1976) . 2-72-3 Flare gas composition froma German refinery (Siegel, 1980) .2-10. ;....2-42-52-62-72-82-92-102-112-122-132-142-152-162-172-183-14-1Characteristics of gases flared in the UnitedStatesCompositionof gases flared during petroleumproduction .Surveyof gases flared fromblast furnaces (Klett andGaleski, 1976) .Surveyof gases flared fromcoke ovens (Klett andGaleski, 1976) . .Surveyof gases flared in the chemical industry (Klettand Galeski, 1976) . .Estimate of the amount of gas flared in the U.S. in 1980Capacityof different flare types ...Properties used inexample of flare designRelative cost of suppressing soot inflares (Klett andGaleski,1976) . ... . ...Properties of fuels flaredSootingtendencies of fuel (Mandell, 1978) .Experimental measurements on flaresRange of concentrations measured inflares studiesExperimental measures of combustionefficiency .Experimental data used in the studyof Becker andYamazaki (1978) withpropane fuel .Estimate of gases flared in the United States ...Experimental parameters and fuel costs for pilot scaleflaretests. . . . .. . . . .xiii2-112-122-142-15. 2-162-182-232-28. . 2-332-352-372-472-562-632-983-24-9LIST OF TABLES (Continued)Table4-27-17-27-37-4Basicflare test matrixFlare head dimensions .Fuel supplyspecificationsTracer usage (ft3/hr S02atInput flowrates .....1%volume in_fuel), .Page4-197-57-77-10. 7-137-5 Time toquench the rate of reactions to 10%of the rateat15000K (Chedai11e and Braud, 1972) . 7-307-6 Sample gas measurements . . . . . . . 7-327-7 Dewpoint (OF) of combustionproducts for propanewith steaminjectionat 11b steam/1b propane 7-347-8 Response of FIDtocarbonatoms in compounds (Beckman, 1970). 7-447-9 Comparisonof techniques tomeasure velocity 7-477-10 Controlled parameters. . . . . 7-487-11 Dependent output parameters 7-59A-1 Comparisonof commercial flare designs (Brzustowski, 1976) A-aA-2 Ground level flares CBrzustowski, 1976) . . . . . . A-13A-3 Suppliers of flare equipment (Brzustowski, 1976) A-15C-1 Flare input andoutput parameters to bemeasured .. C-3C-2 Continuous gas analysis instruments . . . . .. . C-5C-3 Tentative goals for precisionand accuracyof measurements . C-7D-1 Fl are emi ss i on factors . D-1Compositionof flare gas and heat content of individualcomponents . . . . . . . . . . . . . . . . . .D-2D-3Divisionof gas streams . .. . .". 0-20-3xivVariable. SymbolAbCECsCLccpCvoDEOFdEFFRGFRf.GgHHVHChICBFKKTLMNTABLE OF NOMENCLATUREMeaningCross-sectional areaCharacteristicjet radiusLocal entrainment coefficientSpreadcoefficientFlammabilitylean limitConcentrationSpecific heat, constant pressureSpecific heat, constant volumeDistanceLocal fuel destructionefficiencyDi 1ution factorDiameter.Global emission, E = l-UFractionof heat released inaflare that is radiatingFederal Republicof GermanyFroude numberEddy generation frequencyJet momentumfluxGravitational constantHeightLowHeatingvalueof gasHydrocarbon (expressed as equivalent methane)Height of stackIncompletelyburned fuelRadiation heat fluxTracer concentration ratioLengthMol ecul ar wei ghtMassMass flowrateRatioof mass flowratesNumber of eddiesxvi--VariableSymbolN.nopQRReRirSSRTtUuvVw\xyzTABLE OF NOMENCLATURE (Cont1d)MeaningNumber of mol esNumber of atoms per moleculeGlobal efficiencybasedonoxygenconsumptionPressureRadiation intensityper unit volumeof productGas constantReynol ds numberRichardson numberRadius or intermediate variableDimensionless coordinateSti rY'ed reactorTemperatureTimeGlobal efficiencyVelocityVolume. Volumetric flowrateVerti cal veloci ty (ref. Brzustowski)Dimensionless downwind coordinateRadial distance fromcenterlineMass fractionAxial distance fromnozzlexviSymbolynEll.lp'" poTABLE OF NOMENCLATURE lCont ld)GREEKSYMBOLSMeaningLocal combustionefficiencybased on CO2Empirical entrainment coefficientEmpirical factor used to predict the amount of steamrequired tosuppres s sootBurning rate parameterEntrainment coefficientThi ckness of jetLocal efficiencybased on pollutantsJet spreading angleKolmogorov lenqth scaleVi scos ityDimensionless streamline coordinateDens i tyMixing cup densityStandarddeviation;-0- Standarderror of xX Stoichiometric coefficientQ Intermittencefactor-fractionof the time theis presentat radial distance XxviiSubscriptSymbolaadCceFfjLopsstunhcwoverbar[JTABLE OF NOMENCLATURE (Cont'd)SUBSCRIPTSMeaningAirAdiabaticCarbonCenterline valuesEddyFlameFuelJetFlame 1engthConditions at nozzleProductsSonicSteamUnburned hydrocarbonWindAmbient conditionsAmbient conditionsConcentrationsxviiil; ,-1.0 SUMMARYThe U.S. Environmental Protection Agency has contractedwith EnergyandEnvironmental Research Corporation toconduct aresearch programwhich willresult in the quantificationof emissions fromand efficiencies of industrialflares. The studyis being conducted infour phases:I - Experimental DesignII/- Design of Test FacilitiesIII - Development of Test FacilitiesIV- Data Collection'andAnalysisThis report provides the results of Phases I and II of the study.The report summarizes the technical literature on the use of flares andreviews available emissionestimates. Technical critiques of past flareefficiencystudies are provided. Mathematical models of flame behavior areexplored, and recommendations for flare flamemodels are made.The parameters affectingflare efficiencyare evaluatedandadetailedexperimental test plan is developed. The designof aflare test facility isprovided, includingdetails on theflare tips, fuel and steamsupply, flowcontrol andmeasurement, emissions sampling and analysis, and data acquisi-tionand processing. AQualityAssurance/QualityControl Programis alsodescribed. R ~ s u l t s of the testing program(Phases III and IV) will be pro-videdinthe project's final report to be issuedat alater date.1.1 Pollutant Emissions fromFlaresA flare is adevice which allows the economic safe disposal of waste gases,.by combusting them. The waste gases are injected into the openair through atipwhich is designed to promote entrainment of the ambient air and provide astable flame withawide range of throughputs in high cross-winds. In order toreduce flame radiation at ground level, the flare tipmust be elevatedand itsheight will be dependent upon flame size (i.e., flare throughput). If thewaste gas has too lowaheating value tosustain "a flame. auxiliary fuel may beadded. Small flares may utilize fans to provide some air premixing beforeinjection, but most large flares are natural draft with optional steaminjectionto promote fuel air mixing. Flares are usedextensively to burn purged and1-1,-waste products fromrefineries, excess productionfromoil wells, vented fromblast furnaces, unusedgas fromcoke ovens and gaseous wastes fromthechemical industry.An estimated 16Mtons/yr of gasbe flared in the U.S. The amountis difficult toestimate because throughputs fluctuate widelywith time andare seldommeasured. The normal, time-average throughput is in the range ofzero to5percent of design capacity, which is exceededonlyduring emergenciesor upsets. The flared gases fall intothree categories (Klett &Galeski, 1976);Lowheating value gas produced in blast furnaces which account for60 percent of the.weight and 19 percent of the heating value of theestimated annual flared gases;Mediumheating value gases produced in coke ovens and in thechemical industry;High heating value gases flared in refineries which account for 18percent of theweight and 32 percent of the heating value of theestimatedannual flared gases.Pollutant emissions fromflares result fromafailure to completelycombustthe flared gases. ' The pollutant species are normallycarbonmonoxide, hydro':'carbon and soot, and total emissions are assessed based upon an estimate of.flare efficiency. The efficiencyof combustionof aflare, which is ameasureof its abilityto destroy the flaredgas,is difficult tomeasure, and consequentlyestimates of pollutant emission indices vary. Estimates of flare efficienciesvarywidely, some are very high, in excess of 99 percent, whereas others rangeas lowas 70 percent (T.A. - Luft 1974) leading to the conclusion that emissionfactors are unknown. If flares were 90 percent efficient, then emissions ofcarbonmonoxide and hydrocarbons would be approximately 12 percent of thoseemitted by all stationarysources. More important are the contributions offlares as localized sources because of their concentration in refineries andsteel plants where theycould be among the'mostsources ofpollutants if the efficiencies are relatively low.Ii can only be.concludedthat pollutant emissions fromelevatedflaresare unknown. This is due to acombinationof uncertainties in the quantityof gases beingflared and their composition together with the uncertainties1-2\-inflare efficiency. Before adecision can be made whether pollutant emissionsfromflares are of concern, an accurate assessment of flare must bemade. Theoretical estimates of flare efficiencycannot be made, emissionmeasurements fromoperating flares are difficult and previous pilot scale studiesare contradictoryor incomplete. Thus, there is aneed for astudytoaccuratelyassess flare efficiencyas a function of:flared gas composition;throughput;flare design"and(steaminjection, etc.);ambient conditions;scale.Data fromthis study canthen be used to provide an accurate assessment ofpollutant emissions fromflares.1.2 Deficiencies in Previous Flare EmissionStudiesThere have been relativelyfewinvestigations reported in the open litera-ture concernedwi th pollutant emi ss i on or effi ci encY'of fl ares.Table 1-1 summarizes the most recent, known studies, eachofwhichaddressed one or more of thefollowing topics:the emissions of incompletely burnedmaterial;the distance required toburn theflared gases;the impact of steaminjection on pollutant emissions;the effect of ambient conditions on pollutant emissions.Although these studies have made valuable contributions to the knowledge offlare performance, none allowan accurate detennination of pollutant emissionsnor do they provide adequate information on the effects of scale or flared gascomposition.A reviewof the previous studies indicates that data acquisition andmanipulation arecommon problems which prevent an accurate assessment offlare efficiency. These problems are discussed belowin fourmainareas,namely:1-3r..:-",--'J~TABLE1-1.PREVIOUSFLAREEMISSIONSTUDIESThroughputFlareEfficiencyInvestigatorFlareTipDesignFlaredGasMBtu/hr%Palmer(1972).0.5"dia.Ethylene0.4-2.1>97.8Lee&Whipple(1981)DiscreteHolesin2"Propane0.396-100dia.cap.Siegel(1980)CommercialDesign"50%H249-178I97->99(27.6"dia.steam)pluslighthydro-carbonsHowesetal(1981)CommercialDesignPropane4491-100(6"dla.airassist).CommercialDesignH.P.NaturalGas28(pertip)>99(3tips@4"dla.).'. '0 Inabilitytocloseamaterial balance. Measurement of soot concentration. Difficulties caused by flare "intermittency". Lackof scalingmethodology.Closureof Material BalanceThe global (overall) efficiencyof aflare flame can be calculated if.theinlet fuel compositionand mass flux is known together withmass flux of allIhydrogenand carboncontaining species of flaredmaterial at some height abovethe flame whereall reaction has ceased. There is more interest in that frac-tion of the fuel flux that becomes air polluting species rather than harmlessCO2and H20. It is usual toconcentrate on the carbon inthe fuel becauseall'of the ultimateair polluting species containit CO, HxCy' soot). Ifthe carbonfraction of all product gas flux species is summed, the result"should equal the carbonfraction of fuel mass flux. This is the usual massbalance concept and is anaccounting checkon the pollutant species measurements.It is easyto state but rather difficult to implement. Of the studies inTable1-1, onlySiegel attempted toclose the mass balance. Generally he was onlyable toaccount for approximatelyhalf of the fuel carbon in the off-gas flux.Siegel statedthat the largest errors were associatedwith the velocitymeasure-ments needed to.determine themass flux. Siegel circumvents the needfor furthermass balance by using "l ocal burnout II efficiency, and showing that errors in the global efficiencyvalues areminor.Amaterial balance requires time averagedconcentration, velocity andtemperaturemeasurements at some plane normal to the mean direction of flow.These measurements are made above the flame when total emissions are beingassessedwhich requires an integration of the species flux across the total jet.The major errors which prevent adequatematerial balance closure are:Material escapesundetected, becauseat the flame extremitiesdilution lowers itsconcentration belowthedetectabilitylimit ofthe analytical equipment;All the species are not measured;The time average velocityis difficult tomeasure inand nearturbulent flames.1-5A tracer inthe fuel can be used toaid in obtaining amass balance byyielding adouble check on the dilution factor in the product gases. However,the useof atracer do'es not eliminate the need for velocitymeasurements indeterminationof mass flux. More details on tracers will be discussed inSec-tion7.7.2.Measurement of Soot Concentration. Soot represents uncombustedfuel carbonwhich should be included in flareflame efficiencycalculations. Siegel (1980) meisured soot concentrations be-. tween 20 and 80 mg/m3in an intentionally smoking flame, estimating that atthose dilution conditions this reducedflare efficiency by 3to4percentagepoints. More recently, Howes, et al (1981) performedsoot measurements in asmoking propane flame. Usingadilutionfactor obtained fromthe CO2concen-tration, the18mg/m3. of soot measured represented adecrease. incombustioneffi-ciencyof 0.4 percent. It should be noted that these local efficiencies are notequivalent toglobal efficiency, since theywere samples collectedat one samplingpoint.Flare IntermittencyThe term"intermittency" essentiallymeans that at one fixed point abovethe flare, the flame is not present all of the time. Even incalmwind condi-tions, the turbulence induced by the combustion processcauses the flame toundulate andappear unsteady. This usuallycauses corresponding fluctuationsinmeasuredquantitites at fixed points above the flame. Using sufficientsampling times provides one means of time averaging data to avoid this inter-mi ttencyeffect. As di scussed inSecti ons 1and 8, one objecti ve of the proposed. experimental plan is todetermine sampling times so that the characteristicsof the flame aremeasured, unmasked by intermittency.ScalingMethodology'The studies listed inTable 1-1didnot provide amethodologywhereby thedata fromthese pilot scale orsmall, plant scaleflares could be used to assessthe emissions fromthe total populationof flares. Amethodology is requiredwhich will allowdata to be obtained economicallyat pilot scale and used todetermine performance of full scale systems. The current state-of-the-artof turbulent flame structure precludes the use of predictive models. Thus the1-6pilotIt must6-experimental planmust provide data whichwill allowthe effects of scale to bedetenninedand in conjuctionwith developing theories of turbulent flame struc-ture will allowextrapolation to full scale.1.3 Technical ApproachAs will be shown in the BackgroundSection, current information on flarecombustion is fragmentary and inconclusive. This programattempts toanswerthesequestions: What are the combustionefficiencies of small flare flames? Howare these efficiencies influenced byoperational parameters, flaredesign, fuel composition and scale? What are the mechanisms of these influences? Howcan the efficiencies of large industrial flares be estimated?A research programwithemphasis inexperimental measurements on ascale flare is themost cost-effective way toapproach these questions.fulfill these requirements: Representativeness - The hardware and operational conditions mustrelate to full scale practice. Data Accuracy- The measurement methods must be developed and verifiedsatisfactorilytoeliminate the uncertainties that plagued p r ~ v i o u sexperiments. Basic Understanding - Experiments must be designed to bring under-standingon the underlying controlling processes that take place inflare flames. Extrapolation - Informationmust be generatedtoextend the applicabi-lityof the small scale data tofull scaleflares.The designof validexperiments on flares must consider the factthat flare flames are different fromother combustion processes, such asenclosedboil'er flames, in that theyare buoyancy dominated, are affectedbyambient air movements, and lose heat toamuch colder environment.It is commonly accepted that ifsufficient air is mixedwith the fueland if the resultant mixture is kept above the reaction temperature,combustionwill go to near 100%completion. However, these two conditions1-7.;....are not necessarily true inflare combustionsystems, particularlyfor the fueleddies that arefromthe mainflame body .. Because of the geometryof the eddies, they tend to be quenchedat ahigher rate than themain flamebody and hence are more likelyto be extinguished beforeall the fuel is burned.The presence of oxygen next to the fuel ;s essential for continuationofcombustion. In aflare flame, air may be entrained into the fuel jet bynatural convection and assisted by forced convection throughair- or stearn-assist. The effectiveness of these mixing processes directlyaffects thecombustion.reaction. If the mixing is not completed before the burning fuelelements are quenched belowthe reaction temperature, the flame will be extin-guished. Therefore, the research programmust develop the basic understandingof the mixing and eddy behavior of flare flames. This may be aided by modelingwhichwill be discussed in more detail later on in this report.The emphasis of the research programwill be themeasurements andcharacterization of emissions and flame structures. It will include thesecons i derat ions:Four flare sizes 3, 6and 12 inches indiameter) will be linearlyscaled replicas of each other andwill include features of commercialflares.Detailedmeasurements will be made throughout and beyond the visible-flame envelope todetermine profiles of temperature and speciesconcentration.Tracers will be injectedand measured toassess air entrainment.Photographywill . recordflame structuresThe experiments will start with the smaller flares todevelopand verifythemeasurement methods. Once the baseline flare behavior is defined, the effectsof operational parameters and scalewill be studied.The experimental test programcan be logicallydivided intofour tasks.The Task 1 objective is generation of adata baseof gross flame parame-ters as afunction of the complete range of all input parameters. This will bea rapidscreening process on all flareto assess themajor effects of fuelrate, wind level, steamrate, andgas composition. The output measurements will1-8; ..:' ." _1be limitedtovisual and photographic observations of fJame length, formandstructure, and sooting tendency. Video recordings can also supPlement the photo-graphic technique. The utilityof this task lies in its identificationof thoseregimes of the original test planthat need greater emphasis in the succeedingtasks.Task 2will be concernedwithdevelopment and verificationof all measure-ment techniques. This canmost effectively be done using the smaller flare sizes.The measurementswill consist of species concentrationmeasurements inand nearthe flame, including atracer. Development of an integrating hood will be in-cluded. Amajor objectiveof this task is verificationof an adequate carbonmass balance to provide confidence in the succeeding task.Task 3will be concernedwith the detailedmeasurements according to thetest planas revised by Task 1. The major effort will be on the smaller sizes,with the knowledge gained indicating the most important test conditions to beused'for the limited number of large size tests.Task 4is ageneral categoryrelated tocontinuous evaluationof test data,development of modeling and scalingparameters anddocumentation. Amore de-. ' : ~tailed breakdownof these tasks is found inSection4.1.4 Report OrganizationThe report describes the background relatedtoflare design, characteris-tics and emissions inSection2. Sections 3and 4discuss the need for futherwork and a technical approach to carrythis work out. Section5gives a reviewof the potential test facilities which may be consideredfor theexperiments.The designobjectives, afacilitydescriptionand the measurement techniques tobe used in assessingflare characteristics aredescribed inSections 6and 7.Section8discusses data analysis and applicationof the information generatedin the experimental program.1-9., .,-2.0 BACKGROUNDThe primaryuse of flares by industry is the safe ventingand combustionof process gases during emergencyor "upset" conditions. Theyare alsoavail-able todispose of much lower flowrate$ of waste gas that occur during normalprocess operation. It is this latter conditionthat prevails most of the timeand is thus of major importance in determination of flare efficiency. Indus-trial flares encompass awide varietyof conditions; both in terms of the typeof installationand operatingconditions. Important factors are gas composi-tion, heatingvalue and percent dilution by inert gases, flowrate, ambientconditions and combustionsuchas steamor forced draft air.These depend on the type of plant and its location suchthat flare designstend to be sitespecific. Consequently, there is awealth6f hardengineeringexperiencewithrespect toflare design and ,adaptation todifferent conditions,but due to the varied nature of the flaring process, there is alackof compre-hensive information in the open literaturewith regard to specificoperationaldetails as theyaffect thewaste gas combustionprocess. There is onlymeagerand often contradictory informationpublished on the potentiallyharmful mate-rials issuing fromindustrial flares; thus there exists an information gapthat must be closed inorder toassess the environmental impact of flaresystems.Withinthis framework, the objective of this present programis todefinean experimental planwhichwill both improve the understandingof flare combus-tion" and provide ameans of estimatingemissions fromflares of various sizesand characteristics. There is agreat deal ofbackground informationwhich isrelevant tothis task, andthe purpose of this section is to review'the impor-tant available literature inorder todefine the scope of the required experi-mental, program. Of particular interest is informationconcerningflare useand the range of gas compositions encountered in the di'fferent industries,flare designs and the range of operatingconditions, and experimental data andmethods available for themodeling and scalingof flare type combustion.In recent years, a number of surveys of flare use have been carriedout,both in the UnitedStates and abroad, in an attempt todefine the'significanceof flare emissions. Results fromsuch studies providebasis for the estima-tion of total gas quantities flared, the range of gas compositions_encountered,2-1\"\..I;....and, by inference, possible emission factors for the different industries.Information is alsoavailable concerning flare types and designmethods,althoughthe nature of the flare combustion process, and the lackof appro-priatemeasurement methods, has precluded the kindof detailed studythatwould permit adefinition of combustionefficiencies or allowquantificationof theeffect of different flare design parameters. Anumber of small andpilot scale studies have, however, been carried out, and these are reviewedto provide insight into the relevance of small scaleexperiments and thepossibilities for data extrapolation. Much of the available data shows, how-ever, that our knowledge of basic combustionphenomena is lacking incertainareas, and that direct transfer of experimental data fromsmall tofull scaleis usuallynot possible. To this end, asimplemathematical modeling approachis requiredtodirectionfor the designof small scale experimentsand ameans of defining scalingcriteria bywhich the data can be extrapolated.Availablemodeling approaches and basic information concerningthe character-istics of large turbulent diffusionflames are reviewed inthis light.2.1 Use of Industrial FlaresMuch of the informationdesign, andoperationof flares has beenreviewed by Klett and Galeski (1976). More recentlythe German SocietyforPetroleumSciences and Coal Chemistry, DGMK, (Program135-01) analyzed the- /' "results of a "questionnaire sent to 31 German refineries. Unpublished resultsfromasent out to California refiners by the California AirResources Board (Metzger andVincent, 1980) has also beenmade available.Most of the informationwhich follows is based on these studies.Flares are designed for the maximumanticipatedgas release caused byprocess upset or emergencyshutdown. These conditions occur infrequentlyandare of relativelyshort duration. A lower level continuous or semi-continuousrelease is caused by leaks inequipment, necessaryventingof aprocess, andpurging of gases duringstart up and shutdown. These flows, while of muchreducedmagnitude compared toemergency use, occur most of the time. Thus, aflaremust be a veryflexible device, capableof high throughput, and sustain-ed operationat a high turn down ratio. For instance, an instantaneous flowrate of 100MSCF/hr may be demandedwhile sustained normal operationoccurs at1/1000of this value.2-2\..-This section reviews the iridustries inwhich" flares. are used, character-izes the gases flared, and estimates the amount of gas flared in thesetries. Unfortunately, the use of flares is largelyuncontrolled and, hence,the flowrates of gases are infrequentlymonitored. Flowrates are onlyoccasionallymeasured sothat the amount of steamrequired to suppress smokingcan be regulated. Rough estimates have been made of the amount of gas flaredin four major industrial operations: oil refining, blast furnaces, coke ovens,and ethylenemanufacturing for 1974. Here we extrapolate this estimate to1980and estimate the amount of gas flared in petroleumproduction and thechemical industry.2.1.1 PetroleumRefiningThe petroleumindustryflares largequantities of gas fromrefineryoperations and productionwells.Table 2-1 has been constructed fromthe data of CaliforniaAir ResourcesBoard Survey. Althoughthe number of initial questionnaires sent out is un-known, the 63 flares referenced by the 21 respondents showverysimilar char-acteristics. Although "emergency" is the primaryuse, continuous use is alsoassumed. Steamis themost universal means of smoke suppression. None ofthe flares are equipped tomeasure flowrate of the flared gas, soannualamounts are onlyestimates. The compositionand heatingvalue varywidely.These ,results are similar to the results fromGerman refinerysurvey(DGMK, 1981) wherein thirty-one West German refineries responded. Figure 2-1shows the flare gas densityvariationfromone of these refineries duringtheperiod in 1978of Siegel's researchthere. The flowrate is that of Siegel'sside streamand not the main flare.Generally, the gas flared inrefineries is not measured and isdifficult toestimate. However, gases flared inrefineries contribute signif-icantly to the total amount of gas flared in the UnitedStates. Therefore,the amount of gas flared inpetroleummusi be estimated, if the totalemissions of incompletelyburnedfuel is to be estimated.In 1974, a survey (Table 2-2) of 11 of the 288 reflneries in the UnitedStates showed that from0.039to2.8 percent of the refinerythroughput wasflared. The average amount flared excluding the highest number was about 0.22-3TABLE 2-1. SURVEY OF CALIFORNIAOIL REFINERY FLARES(CALIFORNIAAIR RESOURCE BOARD, 1980)FlareSmokeAnnualSteamRefineryFlareDiameterService Flowrafe Fuel~Type(i n)Suppressionscf/yr1 Elevated. 30 Steam Emergency --- ---...a. 351 Elevated 24 SteamEmergency--- ---...a. 351 Elevated 24 Steam Emergency-- ---...a. 352 Elevated---Steam Continuous--- -- ---3 Elevated 30 Steam Emergency---H2' CO. N2 ---C1-C33 Elevated 30 Steam . Emergency--- ------ .3 Elevated 8 Steam Emergency--- ------4 Ground---Steam Emergency --- ---0.404 Elevated 30 Steam Emergency------0.384 Elevated 8 Steam Emergency---Cl C ~ , S 0.304 Elevated 10 SteamEmergency---LPG---5 Ground---SteamEmergency 180M--- ---5 Elevated---Steam Emergency------ ---5 Elevated 16 Steam Emergency--- ------5 Elevated 20 Steam Emergency--- --- ---5 Elevated 10 Steam Emergency--- ------6 Elevated 30 Steam Emergency 36M--- ---7 Ground--Venturi Emergency--- ------8 Elevated 8 SteamCanI t &Erner.--- ---...a.59 Elevated 8 Steam.. Ernergency SOH H'! . Ct -C{i , 1.7H109 Elevated Il SteUl Emergency 3.5M H2C1-C6 , 1.7H2O10 Elevated--Steam Emergency 0.9M---0.5- ..11 Elevated 36 Steam Emergency -------...a. 311 Elevated 36 Steam Emergency--- ---....0.311 Elevated 36 Steam Emergency--- ---....0.311 Elevated 10 Steam Emergency--- ---....0.312 Elevated 18 -Steam Emergency 547M HC, H2S, 1J.3 ->1.RSR13 Elevated 31 Steam Emergency---HC, H2S, 0.33RSR14 Elevated 6 Steam Can't &Erner. 3.9M C I - C 5 ~ H20.4315 Elevated 48 Steam Emergency 111---p.3-0.415 Elevated 48 Steam Emergency 283---p.3-0.415--3D ForcedDraft .Emergency 1.2--- ---15 Elevated 16 Steam Emergency 27.6---p.2-0.3516 Elevated 36 Steam Emergency---Cl .C2, H2 ....0-2.H2, H2. 02.CO,16 Elevated 36 Steam Emergency--- Cl.C2. H2' ....0-2.H2 H2. 02.CO16 Ground---Steam Emergency---Cl,C2. H2, ....0-2.N2. H2,. 02.CO216 Elevated 36 Steam Emergency---Cl ,C2' H2 "'0-2.N2. H2. 02, CO216 Elevated 36 Steam Emergency---Cl,C2, H2 ....0-2.Nz H2. 02, COz2-4TABLE 2-1. SURVEY OF CALIFORNIAOIL REFINERY FLARES (CONTINUED)(CALIFORNIAAIR RESOURCE BOARD, 1980). FlareFlareSmokeAnnualSteamRefineryTypeDiameterSuppressionService Flowrate Fuel(in)scf/yrruel16 Elevated 42 Steam Emergency---ci.C2 H2."-0-2.N2 H2. 02. CO216 Elevated 42 SteamEmergency---CloC2. H2.'\00-2.N2H2. 02.C0216 Elevated 48 SteamEmergency---Cl.C2. H2'I.oQ-2.N2.H2. 2.C0216 Elevated---SteamEmergency ---Cl .C2. H2'\00-2.N2 H2.. 02.C0216 Elevated 70 Steam Emergency---Cl.C2 H2 '\00-2.N2.H2. 2.C0216 Elevated--Steam Emergency--- CloC2. H2.'\00-2.~ N2.H2. 2,C0216 Elevated Steam Emergency Cl .C2. H2.\'\00-2.--- ---N2.H2. 2.C0217 Ground---Steam Emergency ,--- CitC2. H2 '\00-2.\N2 H2. 02.C0217 Elevated 42/100 Steam Emergency---Cl .C2. H2'\00-2.N2 H2. 02. .C0217 Elevated 36 Steam Emergency---Cl .C2 H2 '\00-2.N2 .H2. 2.C0217 Elevated---Steam Emergency---Cl,C2 H2. "-0-2.N2.H2. 02.C0217 Elevated 48/72 Steam Emergency--- CloC2. H2.'\00-2.,N2 H2. 02.C0217 Elevated 12 Steam Emergency---ci .C2. H2 'I.oQ-2.N2 H2. 02. C0217Elevated 12 Steam Emergency---Cl.C2 H2 'I.oQ-2.N2 H2. 02.C0218 Elevated---Steam Emergency}Cl-Cs.NH3.C02'\00.318 Elevated---Steam Emergency10.740H2S 0.319 Elevated 42 Steam Emergency--- ---'\00.319 Elevated 36 Steam Emergency--- --'I.oQ.319 Ground--- ---Emergency--- ---020 Ground---Steam Emergency--- ---20 Elevated---Steam Emergellcy --- ---20 Elevated---Steam Emergency------20 Elevated---Steam Emergency --- ---20 Elevated---Venturi Emergency --- ---21 Ground---Self- Emergency 0.25M Cl -C 30Inspiration2-5r.."L.I:.......'"Q)~2-a .....QJ~1-G::~G::-,.'IJI(Y)2I11II~I1III~IIIII>,I1II.~1.5:1II1(a)0I'1I1I:i1IIIIIQJIIII1~IIIII"-0.5I1II3IIIIII.1II"II,IIIIIIIIIIIIIIIItII1IIIITest~o.~~;~1~6117NI0\(b)NI0)Figure2-1.(a)VariationingasdensityflaredfromaGermanRefinery;(b)ActualflowrateoftestflareusedbySiegel(1980).t..:'TABLE2-2.GASFLAREDINU.S.REFINERIES(KLETTANDGALESKI,1976)NI-....JTotalRefinerytComposition(t)Hydro-CompositionRefineryThroughputFlaredClC2C3[4[5AromatlcOleffnsPilraflnscarbonH2112SINH3Otherbbl/cdlb/cd.154,4370.17001.21.02.40005.45.409.4085.12167,6580.55449.414.512.99.311.1010.387.097.31.90.8003213,0000.1432.35.170.530.540.01l.529.557.498.41.60.000473,7000.14547.914.914.212.28.9-010.387.798.10.60.9005106.0640.05911.126.01.82.82.5012.731.444.2,.043.81.39.86255,0000.0397.832.429.214.36.81.217.775.093.83.22.90(j7239,4000.0568.58.434.441.74.509.187.191.51.41.1008369,5000.2101.39.153.65.38.8010.275.185.30.42.1012.29112,6520.60420.917.834.911.58.6019.774.093.11.61.303.310162,9080.14222.932.118.11.28.9011.877.489.14.66.300II145,0600.18924.213.267.31.14.20094.394.30.40.504.7Total1,899,4190.1923.314.431.611.411.01.012.976.390.21.52.405.812306,5902.788.36.548.433.6.3.100100.0100.00000percent. This number is the same as that estimated by Siegel (1980) for Ger-man refineries. About 12.3x106BBL/cd (cd=calendar day) of petroleumwasrefined in 1974. Therefore, petroleumrefineries flared approximately7.4x106lb/cd of gas based on barrels refined. The figures for 1980are 18x106BBL/cd refined and 10.8x106lb/cdof gas flared.The compositionof gas flared in refineries varies widely, bothwithin arefinery (Figure 2-1 and Figure 2-2 andTable.2-3) andrefineries.(Table 2-1 and 2-2). The amount of gas flared inaGerman refineryvaried byafactor of 22, the density by afactor of 3.4, and the compositionof somespecies by afactor of 5. However, most of the refinerygases flares arelight paraffinic hydrocarbons with large amounts of C3and C4compounds. Anaverage composition for arefinerygas is shown inTable 2-4.2.1. 2 PetroleumProductionGas flared during production of petroleumalsocontributes to the totalamount of gas flared in the UnitedStates. In the past, large amounts of lowmolecular weight gases have been flared fromoil producingwells. This prac-tice has been reduced recently, since the gas product is nowvaluable and muchof it can be sold.The amount' of gas flared in petroleumproduction has not been previouslyestimated, andit is difficult tomake such an estimate. In one report approx-imately0.7percent of theoil productionwas flared (Minkkinen, 1981). Assum-ing aratioof 0.5 percent gas flared toallproduction inthe UnitedStates, the 10produced in the UnitedStates in 1980resulted in the of approximately3 Mtons/yr of gases fromoil production.The compositionof gases flared during productionof petroleumis thesame asnatural gas. These gases aremostlymethanewith small quantitiesof other light hydrocarbongases and inert gases (Table"2-5).2.1.3 Blast Furnaces. \Another major use of flares is todispose of waste gas fromthe blastfurnaces used in the ironand steel industries. As inrefineries, blast fur-nace gftses are flared intermittentlytocontrol process pressures. Gases.flared fromblast furnaces account for approximately60 percent of theweight2-8),-(a)C4HlOC H1ZCH4CZH6CZH4CZHZC3HaC3H6n n3I IIIIIIIIIIIIIII IIIIII-- II IZI.IIIII

97.8Siegel(1980)27.6ReflnllrybVarIable49-178-15290-6401VV13-32V.JVVVVVV97->99Gas\lowes(1981)6C3"840-604413-17(1)VV20.J.J.J.JV.JVNO92-100"owes(1981)3x4"C"480B48(2);NONONRVVVVV.JVNO~99.9~I~(1)AirAssIsted(2)"IghPressureaNotReportedb45-69percentHydrogen,balanceCI-C4hydrocarbons(Table2.3)The characteristics of the flame depend both on fuel and operating.con-ditions, and considerablework has been carriedout todefine flame length,spread angle, andentrainment rate as a f u n t t i ~ n of thesevariables. However,verylittlework is available on the movement of the flame, and the formationand transport of eddies.Flame lengths have been defined in several ways, visible flame lengthbeing the most common. However, it does not represent the end of chemicalreactions which may continue beyond the visibleflame envelope. Therefore,amore precise definitionof flame length is requiredand should ideally bedefinedas the point where the oxidationof the fuel and intermediates stop.However, inthe experiments describedlater, flame lengthusuallyrefers tothe visible flame envelope ... Unburnedfuel, partiallyoxidizedfuel, cracked and polymerizedproducts,and soot may all potentiallybe emitted fromflares. All are products of in-complete combustion. The extent of incompletecombustion depends largelyonthe rate and extent of fuel-air mixing and the flame temperature achieved andmaintained.Jet diffusion flames are not steadystateprocesses. The flame structuremoves aroundas aresult of combustion reactions, buoyancyforces, and combus-tion inducedturbulence. Eddies are formed inthe shear layers of the flameand may breakaway fromthemain body. In some instances aflame eddymay bequenched belowthe reaction temperature and extinguished. This can thenresult inthe productionof incompletelyburnedmaterial whichescapes aspollutants ..2.4.2 Characteristics of Previous Experimental Studies on FlaresA number of previous studies have contrjbutedtothe current stateofknowledge of flare flames. This section reviews the experimental flaresystems andoperatingconditions used in the previous studies.Siegel (1980) made the onlycomprehensivestudyof acommercial flaresystem. He studiedburningof refinerygas onacommercial flare head (typeFS-6-anti-pollutant) manufactured by Flaregas Co. Table 2-3shows thecompo-. .sitions of" the flare gases used which consistedprimarilyof hydrogen (45.4to 69.3 percent by volume) and the light paraffins (methane to butane).2-48 Traces of H2S were also present in some runs. The f1ar.e was operated from0.13 to2.9metric tons/hour (287to 6393 1b/hr). Hence, themaximumheatrelease rate was approximately 235 x106Btu/hr. However, most of the exper-iments were conducted between 49 x106and 178x106Btu/hr.Palmer (1972) experimentedwithaone-half inch IDflare head, the tipof which was locatedfour feet fromthe ground. Ethylene was flaredat 50to 250ft/sec at the exit (0:4x106to 2.1 x106Btu/hr). Heliumwas addedto the ethyleneas atracer at 1to3volume percent and theeffect of steaminjection was investigated in some experiments.Lee and Whipple (1981) studieda bench-scale propaneflare. 'The flarehead was two inches in diameter withone 13/16inchcenter hole surrounded bytwo rings of sixteen 1/8-inchholes, and two rings of sixteen 3/16-inch holes.This configuration had an openarea of 57.1 percent. One hundred and thirty-one (131) CFHof'propane,with 12 CFHof helium, was fired through theflare head. The velocitythroughthe head was approximately3 ft/sec, andthe, heating rate was 0.3MBtu/hr. The effects of steamand crosswindwerenot investigated in this study.Howes, et al,studiedflares producedon two ,types ,of commercial flareheads at John Zink's test facility. The commercial flare heads were an LHair assisted head and an LRGO (Linear Relief Gas,Oxidizer) headmanufacturedby John Zink Co. Since bothdesigns are proprietary, detailed configurationsof the flare heads are unavailable. The LH flare burned 23001b/hr of commer-cial propane on a6-inchdiameter gas pipe. The exit gas velocity based onthe pipe diameter was 27 ft/secwithout air assist and the firing ratewas44 x106Btu/hr withair assist, and the combinedvelocityranged from40 to60ft/sec. The LRGOflare consistedof th;ee burner heads 3 feet apart. Thecombined three burners fired 42001bs/hr of natural gas. This corresponds toafiring rate of 83.7xl06Btu/hr. Steamwas not usedfor either flare, butthe LH flare headwas in some trials assisted byaforced draft fan.2.4.3 The Structureof Flare FlamesDestructionefficiencyand pollutant formation in flares is dependentupon their structure. Inmany previous studies, physical structure was oftenused to flare flames. However, these studies were interested2-49only in the gross properties of the flame, e.g., flame length, flame spread,and the rateofair entrainment. Although these properties are important, amore complete characterizationis required if themechanisms of pollutant for-Imationare to be understood and emissions fromflares assessed accurately.An accurate characterizationof flare flames requires aknowledge of thespatial distributionof temperature, velocity, and species concentration.The time-averagedvalue of these parameters is affected by.theflamestructure since their" instantaneous values fluctuate withtime: The instan-taneous temperature profile influences the rate of destructionofmaterial atanyposition inthe flame. The concentrationprofiledefines the local aver-age species concentrations., the average rateof destructionor productionofspecies, and the geometric limits of burning. The instantaneous velocityprofile describes the flowrate and hence the flux of material in the flame.While some previous investigators have studiedthe temperature, concentrationand velocityprofiles inflares, mosthave concentratedonlyonobservations of overal.l flame length, formand appearance. Observers reportthat the flare flame is typicallynot stationary, and 'that slight changes in,for example, thewind speedanddirectioncancause the flame toshift posi-tionandalter its shape. Many authors also report the presenceof largecoherent structures within such flames, where eddies developalong the lengthof flames and seriesof detached flame pockets results (Figure 2-7a). Eddieswhich are formed on theedges of the flame could be quenched and extinguishedIas aresult of rapid heat transfer to the surroundings and thus be responsiblefor producingmuch of the incompletely burnedfuel inflares. Yet the forma-tion, separation, and quenchingof eddies has been inadequatelyaddressed inprevious studies. However, some studies have recognizedthat quenchededdiescouldcontributesignificantlytoemissions fromflares. Gunther and Lenze(1972) have estimatedthat quenchededdies might result inas much as onepercent of the fuel remaining unburned. Howes, et al, notedthat more eddieswere shed in highwinds and that these eddies contained burningfuel. Leeand Whipple (1981) concludedthat quenchingof eddies was themajor sourceof emissions fromtheir flares. Incontrast, Siegel identifiededdies sepa-rating fromhis flare but concludedtheywere an unimportant source of emis-sions.2-50J tFigure 2-9shows thepotential emissions fromincompletelycombustededdies as afunction of the rateof eddy generation and efficiencyof eddycombustionfor the flare presentedas an example inSection2.3.1. Forexample, the inefficiencyof the flame would be on the order of 1-5percentif the eddy generation rate is 5-20per second, andif the combustioneffici-ency is 50 percent. (Theseestimates are consistent withthose made byGUnther and Lenze, 1972.) These levels cannot be dismissed as small becauseeven one percent inefficiencyis significant whenmost studies have concludedthat flares are greater than 99 percent efficient.The amount of unburnedmaterial is predicted to increaseas the diameterof the nozzle increases, provided the frequencyof generationremains con-stant. However, the frequency of generationis probablysmaller for alargernozzle. Therefore, theseeffects of nozzle size are expectedtocancel par-tially, but not completely. The effects of, nozzle sizewill be discussed inmore detail inSection 2.5.2.4.3.1 Temperature Profiles in FlaresThe spatial distributionof temperature inaflare flame is affected bywind, steaminjection, and the heating value of the flared gas. With lowwind velocities, the temperature profile inaplane above the flame appearsto be normallydistributed (Figure 2-10and Siegel, 1980). However, recentevidence producedat Caltech (Oimotakis, 1981) indicates that the profiles inaturbulent flame are relativelyflat except near the edges. The normalradial distributionof properties, usuallymeasured inturbulent flames, isan artifact caused by fluctuations of the flame about the axis, i.e., theflame isless frequentlyat large radial distances at the centerline. Consequently, the time-averaged temperature appears todecreasewith,increasing radius.The time-averaged temperatures measured in such systems are not usefulfor making estimates of combustionrates. Inmany instances, the averagemeasured temperatures are less than 600Cwhich is near the ignitionture of most hydrocarbon fuels (Table 2-13). Hence, much of the fuel wouldbe quenched before being burnedcompletelyif the average temperature wererepresentative of the combustionprocess. these temperatures arenot representative of the heat release zones, but. rather the average tempera-2-51EddySize ~ FlareTipDiameter16 in. Diameter Flare25%EfficiencyofEddy CombustionuQ)III--%unburnedfuel emittedFigure 2-9. Estimates of flare emissions due to incompletecombustion of eddies.2-522-53tureat apoint where hot combustingeddies move throughat some frequencyfollowed bycoldeddies. Thus, while the instantaneous temperatures of burn-ingeddies can be used to predict combustionefficiency, the averagemeasuredtemperature cannot.The averagemeasured temperature profiles in aflame help todefine theboundaryof combustion. Measured temperatures which approachambient onesindicate that fewhot eddies are passing themeasuring locationand, therefore,the location is outside the nominal edge of the f l ~ m e . Axial temperature pro-files in Figure 2-10for twodifferent flowrates reported by Siegel (1980) in-dicate that increasing the flowrate lengthened the flame and raised the temper-ature at the same distance above the nozzle. The radial temperature profiles,showthat at apositionof 6.5meters above anozzle firing 3800 lb/hr of fuelthe fl arne is appr,oximately6 meters wi de. Thi s fl ame is much wi der than pre-dicted byjet theory. However, Siegel used adivergent flare head which wouldinduce more rapid spread of thejet than a normal head. However, others (Leeand Whipple, 1980) who have used normal flare tips have also found more rapidspread of flare flames than predicted byjet theory.2.4.3.2 ConcentrationProfiles in FlaresThe major components inaflare flame are O2, CO, CO2, hydrocarbons,H20, and carbonaceous particulates (soot). ' Since the flare entrains ambientair along its flowpath, the concentrationof combustionproducts becomesincreasinglydilute. Thus, lowconcentrations of incompletelyburned fuelspecies are not necessarily indicative of anefficient flare. Rather, theefficiencyof the flare is the product of the instantaneous concentrationand the instantaneous mass' flux surrmed over an envelope enclosing the flare.Inaddition, the valuesmust be averagedover asufficient time period toensure that the emission ratedoes not varywithtime. This has not beendone in previous studies and is impossible if conventional sampling tech-niques are used.Themeasured normal distributionof concentrations inaflare flameappear to be the result of flame fluctuations. However, the averaged con-centrationprofiles of the flame aremore valuable than the average tempera-ture profiles. The average concentration profiles can be used tocalculatelocal destruction efficiencies, identifythe regions of average production2-54or destruction of aspecies, specifythe concentrations of incompletelyburnedspecies, as well as to indicate the flame envelope.Concentrationmeasurements inflare flames have beenmade by Palmer(197i), Siegel (1980), Lee andWhipple (1981), and Howes, et al (1981). Table2-16 shows the species and concentrations measured in these studies. Palmermade onlysingle-point measurements of heliumandethylene. For inlet helium/concentrations of about 1percent, the concentrationof heliumat the end ofthe flame was between 21 and 206 ppm.' The measured range for hydrocarbons wasfrom0to 104 ppm. Lee and Whipplemeasured CO, total hydrocarbons, propane,and heliumconcentrations inapropane flame dopedwith8percent helium. Be-,tween 50 and 81 inches above the flare nozzle, the concentration ranges of thespecies were: CO, 7to 1000 ppm; total hydrocarbons, 0.3 to 270 (arbitraryscale); propane, 1to6000 ppm; helium, 130to 2700 ppm.Measurement of lowconcentrations of, species does not ensure that emis-sions fromflares are insignificant. The concentrationof species escapingthe flame are diluted byair andexit the flame regionat highflowratesthrough alarge area. Figure 2-11 estimates the relation betweenmeasuredconcentrationof 'propane andefficiencyat v a r i o ~ s dilution levels and Figure,2-12 estimates theeffect for emissions of carbonmonoxide. At the estimateddilution of 1000, 30 ppmpropane or 10 ppmcarbonmonoxide represent aonepercent loss of efficiency.Siegel has takensufficient data to showthe variation in average concen-trationdistributions withdistance under no-windconditions. Figure 2-13shows themeasured O2, CO2and gaseous hydrocarbons along the flare axis fortwo firing rates. These concentrations conformtoone's expectation. Fuelburnout is displaceddownstreamfor the flare withthe higher firing rate.For boththroughputs, the concentrationof fuel decreases more rapidlythanthe CO concentrationwhich indicates that the concentrationof fuel is decrea-sing byconsumptionas well as bydilution. However, the instrument used bySiegel tomeasure COwas not sufficientlysensitivetodetermine the COcon-centrations tothe required accuracy. Extrapolation of the fuel concentra-tions tozeroyields areasonable estimateof flame lengths. The 780 kg/hrflame is 6 meters long and the 1100kg/hr flame is 8 meters long.2-55~TABLE2-16.RANGEOFCONCENTRATIONSMEASUREDINFLARESSTUDIESCONCENTRATIONPPMVTracerHCCOC3H8CO2.02SootAuthor~v-!!!.v!!!.v!!!.v%%mg/m3Palmer21-2060-104Lee&Whipple130-2700a7-10001.5-60000.5-1.019.8-20.3Siegel-0-12000-2000-0.5-316-2023-81NIHowes0->10000-50000->1000.0.1-7.29.6-20,u ~.~88u'r-It-864-w~ o84'r-N.fJIIIICJ1:::s82........0 E 0u8078767472101001000Concentrationascarbon,ppmv10,000100,000Figure2-11.Effectofpropaneemissionsoncombustionefficiency.Propaneasfue1~r10098969492........;,.,g--.....-90>.u c88(lJ'r-U'r-4-864-wN~84IU1'r-oo+.lIII~82..0.E 0u80787674721.0D=DilutionFactor'10100100010,000Concentrationascarbon.ppmvFigure2-12.EffectofCOemissionsoncombustionefficiency.COasfuel.Siegelo 780 kg/hr.1100 kg/hr no steam541o20o700900800300200.1,000. 100600 12 3::=::, Q..Q..~500 10 CO2%~:r:xu400 8 22 3 4 5 6 7 8Height Above FlareTip, MFigure 2-13. The effect of throughput on concentration profilesin two flare flames. (Siegel,1980).2-59Radial profiles of concentrationare also useful for definingthe flameenvelope. Profiles reported by Siegel (Figure 2-14) showsome unusual charac-teristics. First, the profiles are asymmetric. The more rapiddecrease offuel, the larger concentration CO2andmore gradual increase in O2concentra-tion on the right side of the center lineall indicate that on the averagethe flame burns more to the right than to the left side of the center line.This could be caused by.distortionof the flare off-axis by the nozzle orwind, or by sampling for atime periodwhich was insufficient toaverage thenormal fluctuations. The concentrationmeasurements also showconsiderablescatter on the left sideof the center linewhich may be caused by intermit--tent sheddingof eddies. Such ,phenomena are not accounted for in the samplingprocess.Finally, fuel does not asymptoticallyapproachzeroconcentrationat theedges of the flame. Instead, the fuel concentrationapproaches an asymptoticvalue of 10 ppm. This concentrationof fuel might contributeas much as 1percent toflare inefficiency. Inaddition, the concentrationof fuel shouldreach an asymptotic value of zeroif transport of fuel occurred by diffusionalone. The finite asymptoticvalue implies that fuel is being transportedfromthe center line by some mechanisminaddition todiffusion. Once again,themost likelyexplanationis transport of fuel in large eddies that havebroken away fromthe flame envelope, aphenomenon noted bySiegel, Howes, eta1, and Lee and Whipple. Siegel concludedemissionfromquenchededdies wasunimportant, while Whipple concludedthat itwas. Althoughthe species con-centrationprofiles of flare flames normallydefine the envelope of the heatrelease zone, interpretationof time-averagedprofile concentrationto produceflame envelopes is made difficult becauseof the fluctuating nature of buoyantturbulent flames.2.4.4 Flare EfficiencyAmajor objective of many experiments on flares has been to determinetheir combustionefficiency. The efficiencyof combustion is difficult tomeasure directlyand, consequently, various methods have been proposed toprovide pollutant emission indices for flares.2-6050 ,.....---..,...----r---............oGas = 1720 Kg/hrSteam=0Kg/hrp=0.59 kg/m3Z =6.5m0.2o..... -+- +-__-+ -+- +-__

20. 0L..L__--L__ __ -3 -2 0 1 2Radial Position, M20.5E0.0.>,::I:XUNou

. NoFigure 2-14. Radial concentration profiles in aflarefl arne. (Siege1, 1980) .2-61Fivemethods have beenused tocalculateflare efficiency inpreviousstudies (Table 2-17). Degree of carbonconversion (e.g., Siegel). Extent of destructionof flaredgas (e.g., Lee, Palmer). Amounts of final combustionproducts formed. Emissionof undesirable products/intermediates. Extent of the oxidation,process.In Siegel's study, efficiencywas based upon local carbonconversion, U,defined in Table 2-17. This requires that measurements be made of the localgas velocityand concentrationof all carbon containing species in order tospecifyan overall efficiency. Siegel attempted toclose a material balance,but he was unable to account for approximately 50percent of the input carbon.This was attributedto errors in thematerial balances caused by inaccuraciesin themeasuredvelocities. While inaccuratelymeasuredvelocities cancon-tribute toerrors in thematerjal balance, other factors must be considered,such as: Improper use of the average velocityand concentration. Errors in the average concentrationcaused by short sampling times. Incompletely burnedmaterial whichescapedthe flare undetected. Use of inappropriate techniques tointegrate the mass fluxes.Howes, et al, studiedthemeasurement methods that may be applicable toflares. Theydetermined species concentrations by analyzingsamples with-drawn fromthe flame usinga heated probe. The extractedsamples were anal-yzedcontinuouslyfor CO2, CO, 02 and THC. The probe was also used toextractgrab samples whichwere analyzed byamodified SASS trainfor particulates andorganicmolecules.Howes, et al, reported local destructionefficiencies between 92 and 100percent, measured2 m above the flame. The lower values were for sootingflares, and the higher values were for flares which suppressed soot. Thestudydid not: (a) report global efficiencies, nor (b) confirmthe accuracyof local efficiencies bydifferent techniques.2-62TABLE 2-17. EXPERIMENTAL'MEASURES OF COMBUSTION EFFICIENCY1. Used bySiegel(gl oba 1)u = _m=.c,_. __ . m1n(CH)c mn, 2. Used by Lee_[ fD] xDF]x 100DE - 1- [C H- U(1 oca 1) . 38 0.where DF=[He] in sample[He] in flare gas3. Used by Palmer DE=(1 oca 1)where DF=1_xDF] x100[C2H4]0He] in sample )[He] assumi ng 75% reacti on13.254.COUsed by Howes! DE = - Mc2( 1oca 1) MC02+MCO+MHCc c c=DE= 1where [HC] isthe concentrationof hydrocarbons as methane.-outmuhc-inmuhc5. Proposed:(global)6. Proposed:(gl oba 1) .-outF =m of final productsmout assuming 100%conversion8. Proposed:(gl oba1)E = 7. Proposed:(global)r-outd . bl .me . un eSlra espec,es- inmca consumedo = 2Theoretical O2required for 100%conversion2-63K=Grams tracer in feedT Grams carbon in feedTABLE 2-17.9. Proposed:(loca 1)EXPERIMENTAL MEASURES OF COMBUSTION EFFICIENCY (CONCLUDED),12KT [C02J ppma = [T] ppm~10. Proposed:(1 oca1)11= 1-(36 [C3Ha]ppm+12 [CO] ppm+ (:&)[TJ ppmM,-where RT/PNconverts to volume units. msootis grams soot insample. c11. Proposed: =12.wheremcdenotes carbon fromthat species.then aI -+ Howes1 DE (# )muhc+ mCO+msootProposed: 111= 1_ c c c(local) mCO2+ to+ uhc + sootc mcmc, mcNote that ifmsoot::::: 0 ,CCOMMENTS:(a) Equations (2) and (3) are incorrect because theyrelate to propaneand ethylenedestructionefficiencyonly, and not tooverall com-bustionefficiency. Soot, CO and lower hydrocarbons mayalso bepresent as uncombusted species.(b) Equation (4) is subject toexperimental error.(c) Equations (5) -'(a) are unworkable incurrent format without ahood,as is Equation (1)(d) Equations (11) and (12) relyon assumed mass balance.SUGGESTI ONS :(a) Use Eq. (9) until mass balance is obtained,then use Eq. (10) toquantifylocal efficiencies.(b) Given 1 0 ~ a l efficierrcies, weight, themaccording toarea andvelocitytogive global efficiencyand check this result withthat obtained by using the hood and Eq. (1) or (7).2-64Palmer (1972) measured the local efficiencyof an ethylene flare. Hisdata are tabulated inTable 2-16. Palmerwithdrewgrab samples into2.25liter bags withaprobe fromaposition two feet downstreamof the "flametip.: The samples were analyzedfor helium(used as a tracer) by gas chroma-tographyand mass spectrometryandfor ethylene (fuel) bygas chromatographyand flame ionizationdetection.The.heliumtracer technique used by Palmer toobtain local combustionefficiencyis questionable. Heliumis apoor tracer because the diffusionvelocity is much higher than themajor components in the flare. Consequently"heliumwill diffusemore rapidly away fromthe flame region and produce dilu-tion factors that are erroneouslyhigh. The sampling time used by Palmer,one minute, is probablyinsufficienttoaccuratelydetermine the species con-centrations of afluctuatlng flame. In addition, Palmer sampled onlyon theaxis and did not attempt tosample or integrate the concentrations toobtainaglobal combustionefficiency. The sampling used by Palmer could notaccuratelydetermine the concentration of high-boiling-point hydrocarbonswhichwere observed. Also, the heliumanalyses, whichwere the basis for themass balance calculations, onlyaccurate to6.7percent. Consequently,Palmer's data are of limitedvalue inassessing theemissionof incompletelyburnedmaterial fromflares.Lee andWhipple (1981) have also reported local efficiencies measuredon apilot-scale flare'using aheliumtracer todeterminedilutionfactors.Destructionprofiles of propane, hydrocarbons, andcarbonmonoxide are shownin Figure 2-15. Although no soot concentrationmeasurements were attempted,the data can be utilizedtodetermine minimumcarbonemissions by consideringthe sumof propani and COonly (Figure 2-16). Theyalsoreported combustionefficiencyof 96 to 100percent. However, the data contained the same poten-tial errors as the other data on flares. Only local concentrations weredetermined. The tracer usedwill potentiallyyield too highadilutionfactorbecause of the highdiffusivity. The sampling timewas minutes, but ascwill be shown later; still too ,short toachieve the accuracyrequired. Itshould be notedthat escape of 0;001 percent of the fuel represents about aone percent loss in efficiency"and this was ignored.2-65C'lo.....Jco::I:MU1.02.03.04.04 04.0rax +.J.r::.#of pOints=Nav.g. (ArittiC'l. ~OJ.r::.m1 n ~ttlC1Jc..U::I:1.0C'l0.....J0(HC)12(CO)-Ec..c..0uC'l1.0.....J0(50) (61) (72) (81)60 70 80Distance Above Burner Tip, Z(in)Figure 2-15. Species centerline concentrations as afunction ofheight above burner. Fuel 92 percent propane,remainder heliumtracer. Calmwind, sootyflame.Effective nozzle diameter 1.51 in.(Lee and Whipple, 1981).2-66CV=0.02Grab Sample 1.5ft.Off AxisN=l JGrab SampleN=3, CV=0.417'N=Number of data pointsCV = Coefficient ofVariation\ON=l\\(U= 0.9990)(U=0.9900).(U= 0. 9000 )..

0-..... COVl::I:Vl ('t").... WEl.LJ OW..0-

l'OWQJ

l'O,....U.:::>I-IIl.LJ.......I ....L.- """ ---l50 60. 70 80Distance Above Burner Z (in)Figure 2-16. Summary of flare emission, excluding soot, asafunction of height above burner tip. Sameconditions as Figure 2-15. (Lee &Whipple, 1980).2-67Concentrationof species has beenmeasuredabove and inflares. How-ever, interpretationof these results isdifficult because of the intermit-tent nature of flare flames. Estimates reveal that in order todeterminethe efficiencyof aflare, concentrations must be measured down toabout 10ppm. Instruments used by Howes, et a1, a n ~ Lee and Whipple were sensitiveenough so that emissions could have beendetermined. However, instrumentsavailable to Siegel were not.2.4.5 Productionof Soot in FlaresThe amount of soot produced inflares canmateriallyaffect the flareefficiency. Figure 2-17 shows that aconcentrationof 15 mg/m3of soot atadilution factor of 1000corresponds 'toa loss of one percent in flareefficiency. However, fewstudies have measuredthe concentrationof soot.This is so partly: because themeasurements are time consuming., Sometimes,soot measurements are consideredunnecessary because some flares burn non-sootingfuels, while others inject steamor air to suppress soot formationduringmost operatingconditions.Howes, etal, and Siegel made limited soot concentrationmeasurements.However, their measurements were neither systematic nor complete. Howesmeasured soot concentrations ranging fromless than 0.1 mg/m3for naturalgas flared on a LRGd head, to 18.3mg/m3for C3H8flared on an ~ H head. Theefficiency loss caused byasoot concentrationof 18.3mg/m3at adilutionfactor of 1000is about 1.5percent. Siegel measured soot concentrationranging from20to 80mg/m3in asootingflare. He estimatedthat productionof this amount of soot resulted inanefficiencyloss of 2to 4percent.Figure 2-17would indicate an efficiencyloss of about 1.5to 5.5percent atadilution factor of 1000. The .observedconcentrationof soot fromflareflames can, therefore, substantiallyalter combustionefficiency. The con-centrationof soot should be measured infuture studies to levels below10mg/m3.2.4.6 The Effect of' Wind on the Performanceof FlaresElevatedflares areexposedto uncertainweather condttions includingchanges inwind velocities. In acalmatmosphere, the flare flame primarilyfluctuates about the center lineof the flare head. As the crosswindvelocity2-68,.10098969492..-...90(j-Q.-..J.>,.88u c QJ......u86......If--If--Nw84IcC"I0~......~82Vl';:'IDF=DilutionFactor..0E800u7876747210/100SootConcentration(mg/m3)100010,000Figure2~17.Effectofsootconcentrationoncombustionefficiency.Propaneasfuel.increases, the flame axis is bent inthe directionof the wind. As the windpromotes penetrationof air intothe flame, combustion is enhancedand theflame is shortened. When thewind velocity increases further, thewindvelocitydominates and the flame becomes longer, and the possibilityforstripping unburnedmaterial fromthe flame edges increases.The effect of crosswind on the efficiencyof flames was studied toalimitedextent by bothSiegel and Howes, et al. Howes, et al, studied flamessubjected towinds from8to 17 mph, while Siegel studied flares subjectedtowinds up to 15 mph. Acrosswindmakes measurements and the closing of mate-rial balances in flames verydifficult. Thus, reportedmeasurements are oflimited value.Global efficiencies have not beenmeasured for flares in the presenceofwind, but some local efficiencies have been reported. For aflare in a6.7m/s wind, Siegel reports that local combustionefficiencyon the center-lineof the flare approaches 97 percent, compared to better than 99 percent at acomparable point inquiescent conditions.2.4.7 The Effect of SteamInjection/Forced Draft on the Performanceof FlaresIn practice several methods are used to reduce soot production and.improve the efficiencyand visual appearance of flares. Thesemethods nor-mallyserve to improveair induction andmixingat the root of the flame.Air can be induced by the use of separate steamjets injected intothe baseof the flame, by aerodynamicmeans suchas in the Coanda flare, bypremixing,~ or bythe use of forced draft fans. -The inductionof additional air intothe flame typicallyincreases combustion intensity, shortens the flame, andimproves the efficiency.Howes, et al, studiedaJohn Zink LH flare which used aforced-draft fantoassist airmixing, and a John Zink LRGOflare which burned highpressuregas. The estimatedlocal efficiencyof the LRGOflare was slightlybetterthan that of the LHflare., But since the LRGOflare burnedmethane and theLH flare burnedpropane, comparisons are difficult. In one case, Howes, etal, estimatedthe local efficiencyof the LH flarewithout using the forced-draft fan. In this test, soot concentration increased by afactor of 20, andthe estimated local efficiencydropped fromgreater than 99 percent to 922-70percent. Use of forced-draft fans definitely improves the efficiencyofflares. However, forceddraft fans are impractical on the very large ele-vatedflares which potentiallycancontribute significantlyto the emissionsof incompletelyburned fuels fromflares.Inductionof air through the injectionof steamis the most commonmethod of improving the efficiency o ~ large elevatedflares, and both PalmerandSiegel have studied theeffect of steaminjection into flares.Palmer estimated the efficiencyof a flare with andwithout injectionof steam. He reported that the flame lengthdid not stronglydepend on exitgas velocities for velocities between 50and 250ft/sec. However, injectionof steamchanged the physical characteristics of the flame. Without steaminjection, the flame was betweenfour and one-half and five and one-half feetlong, "lazy," reddish, and smoky. When 0.3 lbs of steamwas injectedforeach pound of ethylene, the flame became turbulent, shortened to two feetIandemitted no smoke.Siegel studiedtheeffect of steaminjection on the performanceof flaresmore systematically. He demonstrated that injectionof steamchanges thephysical characterist)cs of the flare and that the changes are reflected inthe concentratiqnprofiles of theflare and the local combustionefficiencies.An optimumlevel of steaminjectionexists, which depends on the characteris-tics of the gas being flared.Steaminjectionchanges the physical characteristics of the flare in amanner which suggests that the intensityof combustion is increased becausethe flame is shortened, its luminosityis reduced, and soot formation issuppressed. Figure 2-18shows that the flames are shortened byabout 50percent at asteam/gas ratioof approximately0.1. Without steam, the flamewas verysooty. Steaminjectionchanges the color of the flame fromorangetoyellow. When the steam/gas ratio .reaches about 0.25, soot productionwassuppressed. At steam/gas ratios of about 0.55, the flame develops ablueinner cone and ayellowouter cone. When the steam/gas ratiowas above 0.55,the blue inner region spread totheentire flame andit resembles afullypremixedflame. Increasing steam/gas ratio past 1.35produces awhite innerconewhichcontains steamand water.2-71.,-~Q).+.JQ)::::-,-'=+-'01t:Q)-JQ)E' 0Ricou and Spalding (1961) continued tothat theoreticalscaling basedon the Froude number, Fr, led to good predictions for thedifferent flames. The Froude number is defined as:where:Un2( ) .!::Se.PPogdoCT======specific heat of gas at constant pressuretemperature (absolute)gravitational accelerationmass of fuel in injected gasheat of combustion of fuelvelocityat nozzle-.,...... -Air entrainment by natural convectionoccurs faster than forced convection(but at differentFigure 2-30shows the assymptotic limits offorced and natural convection. Figure 2-31 shows the correlation betweentheoretical and experimental values scaled by Froude number.A studyof aturbulent flame in acrossflowwas done by Escudier (1972).Escudier proposedthat fluid is entrained intothe plume at arate propor-tional to the velocitydifference in boththehorizontal and vertical direc-tions. Thus, referring to Figure 2-32, the mass conservationequationmaybe written:2-89, ,1'-) Asymtope for pure, natural convection,'I'II /1 //Y ,

/1'/for pureforced convectiono 1"'-__.L' __ __ __--"'__1 0 10 :100,xI- - F--do Po r'10_INI 10E.......E-Figure 2-30. Theoretical predictionof entrainment inbuoyant jets. '(Ricou&Spalding, 1961;reprintedwith permissionof CambridgeUniversity Press).2-90,-105210 I(;,5

0E:-002E:-001(;,05()'02heo. .I 1.1I I. ) retialcurve" (figure2 I IYYih's dam_-' /r I7"'":--,."I IJ"'i'/Qe:ves' md Bociter's-. /dam V ./.

, 6/ aJ"o\6

" . .......;;i/!"-r-mF,-t.o-32 Z(1'1) i F.-I- 1" - - 1-oJmOPor-

.r-1/O{) 1 002 ()oQ 5 Qol 0'2 ()05 10 2:0 G:) 5 10Figure 2-31. Entrainment by buyoant jets and flames.+, unburnt propanejet; 0, unburnt hydrogenjet;0, pre-mixedair-hydrogenflame; hydrogendiffusionflame; e, propane diffusionflame(RicouandSpalding, 1961; reprintedwithpermissionof Cambridge UniversityPress).2-91zUcow.J~ gr...,w.-----------__ x, .: ,-.Figure2-32. Definitions of x, z, the curvilinear coordinate ~ ,horizontal and vertical velocitycomponents u and w.velocity U, and flame radius o. The gas jet dischargesat velocity Wj with densityPj fromanozzle of diameter d.intoacross streamof density P at velocity u. Jco co2-92"-.where y,are both. empirical constants. Brzustowski, et al (1975) noticedthat for Escudier's equationto reduce to Ricou and Spalding's without cross-wind and buoyancy, ymust have aspecific value. Thus, we get=2'Pmo[0.08 ( P: )"(u- +"Ewhich assumes that we canextendapplication of the equation back to the nozzle, .=0) withminor errors.Brzustowski has also studi edjets in'across wi nd. He defi nes aburni ngrate parameter Bby(4)M ) =-0.233e pp dm Mo0where Mp' Moare molecular weights of products and reaftants,arestoichiometric coefficients.Bvaries from0for.a coldjet to 1forajet whose oxidationrate is limited by the entrainment rate. The generalsolutionof these equations cannot be expressed inclosed form. However,closed formsolutions can be obtained for the casewhere (lE=0; i.e., buoy-ancy is negligiblysmall, and entrainment is given by the Ricou and Spaldingequation. These solutions can be qualitativelycomparedwith real flames topredict that high entrainment will causea flare to bend over earlier. Thelimiting solutions are:m*m/mk: -(a) = = 1+0.32 (p""I po) 2= 1+0.32

0(b) Yp=-.0746(4)pMpl4>oMo)S =mass fraction of products1 + 0.32t(c) Yf=1Ym*p2-93whereE,; =

(:: -d-oand(d) x::X(:: )'=

.10.32Nel'lr tI":e nozzle, this may be appr.cximatedbyz = Define the flame lengthas the location beyondwhich mass flowof combus-tion products stays constant, thenfL =13.4 +,sinceat that point Yf=O. Note that at this locationm*,(fL)=1+ 10.233Under the above assumptions, this gives the dilutionfactor for acase when nofuel is burned. This model is then consistent when obtainedfromthe flame-lengthequation, leads totheobserveddilutionfactor inthe equationfor en-trainment.As an example, consider the combustionof propanewhichoccurs via'In this case, the following values are specified,where'M is the averagemolecularweight of the products. For propaneatp ,ambient temperature and pressure, is 1.24and leads toaflamelengthpredictionof 60/Sregardless of crossf1ow. The lengthof a turbu-lent propanediffusionflame has been found to be about 300dj, implying2-94s = 0.2 (Hottel and Hawthorne, 1949). This derivation is applicable onlywhen the Froude number is large enough to neglect buoyancyeffects.In addition to presenting the limitedcase of aflame in the absenceof buoyancy, Brzustowski presentedanumerical solutionto the 13differen-tial equations describingthe general diffusion flame system. The parameterSwas evaluatedfor the no-crosswindconditions bymatchingpredictedandobserved flame length andwas found to be about 0.15. The empiri.cal entrain-ment coefficient, a, was obtainedwith = 0.05at nozzle velocities of15m/sec and 57m/sec. Avalue of a = 0.13matchedthe data quitewell.Compositionprofiles are presented in Figure 2-33. Note that fuel andoxygen can exist together inthis model since S does not have to be unity.The 13equations solvedare: continuityof fuel, intermediates, oxygen,products, and nitrogen; energy balance; horizontal and vertical momentumbalance; total continuity; ideal gas equationof state; defin-itionof S; and acalculation' of residence time. The method used for solvingthe systemwas not given.Becker and Yamazaki (1978) presentedtheir data for turbulent flames inadifferent form. They used the Richardson number which corresponded to thereciprocal of the Froude number.. Theyalso .definedanear-fieldcoordinateand afar-field coordinate.Richardson number: Ri =9do/Uo2Near-field coordinate:=Ril/3z/dFar-fieldcoordinate: . ="Ril/2Theyalsodefinedacharacteristicjet area with radius (b) suchthat at theradial position2pu, .': ,-.whereis the average vertical yelocityat b and values with subscript carecenter linevalues. Thejet radius, b, may be expressedas afunction ofaxial distance (z) fromthe nozzlewithaspreadcoefficient, Cs.b = C zs2-95.........a......>-25 50p75 100 1 ~ 5 200 300 400~ L Fi gure 2-33. Computed prof; lesof compos i ti on and dens ityfor d =0.005m, w= 22.1m/sec, and u ~ = 2.55m/sec.(Note scale changeat the end of the flame).2-96They found that up toof about C is constant at about 0.071. Beyonds . Csperhaps due tofree convection buoyancy. Other deriva-tions implied a Cslimit of 0.07 for forced convection and 0.10for free con-vection. It is.interestingt6notethat the flame jet appearedtoemanatefromavirtual originabout five diameters above the nozzle tips,whilesimplejettheorywould predict avirtual.originof about three diametersbelowthe nozzle This may be due to boundarylayer effects of the sim-ple tube used in the study.In contrast with Ricou andBecker and Yamazaki define adif-ferent entrainment coefficient CE such thatdm= C (7TpG/4) 1/2dz Ewhere mandGare thejet momentumfluxesp is the II mixing CUpli density (stirredreactor).It was found that forof 2and pof the entrainment was 0.36. This value is different fromthe 0.16found by Ricou and Spalding(Figure 2-31).the correct model is somewhat uncertain. Table 2-18sunvnarizes the data used by Becker and Yamazaki. Further discrepancies areevident when the free convection limit of 1.84is comparedwithexperimentalresults shown in Figure 2-31 which suggest acoefficient of about 0.60.The studies discussed previouslydealt withthe aerodynamics and reac-tions of fuel jets.(1980) developeda systemof ninedifferentialequations to describe soot formation. Some of his results are presentedforan ethylene flame in Figure 2-34. Magnussen developed three kinetic rateexpressions based on and final products and used the onewhichwould limit the reactions under the specific conditions. Without adefinite numerical solutionwould have onlylimited use and is probablybeyond the scopeof astudyof flares.Inresults fromthe past modeling studies of burningjetsarenot consistent. The most recent work (Becker) contradicts theearlyresultsof Ricou and Spaldingand does not suggest amethod of reconciling the dif-ference. the limited case solutions of Brzustowski 1977) may be used to predict the approximate nature of the flare.2-97...,,~.NIU)ex>TABLE2-18.EXPERIMENTALDATAUSEDINTHESTUDYOFBEKCERANDYAMAZAKI(1978).WITHPROPANEFUEL(REPRINTEDWITHPERMISSIONOFTHECOMBUSTIONINSTITUTEMassFraction--Far-FieldNozzleNozzleGasofSource105RiLengthSteamFlameMeterVelocityMaterialNozzleNozzleDia.CoordinateEnddoReo0LIdo~LReLu.w0mmm/s'5.633.50.0602131044312821.0-53005.637.80.0602291090.416015.59400.5.6717.50.0597657018.119511.0175005.6927.10.0594102007.612189.24243005.7142.00.0590158003.192467.80353005.7259.20.0588223001.612656.69437003.2576.90.0582164000.55.2784.91306007.9191.40.0602482000.912896.03860001.0..----......----,.----...,..------.oo~~ M E---s::. ~ 0.5+-'ltls..+-'s::OJus::O'u.pooVlox- axial distanced- nozzle diameter100x/d"-Figure 2,,:,34. Experimental Soot Concentrations on the Axis of the C2H2Diffusion Flame (Re=7000) Compared with Predictions(Magnussen, 1980).2-992.5.3 Solutions of theTransport EquationsAgeneral method of obtaining numerical solutions to the various partialdifferential equations (POE's) governing the flare systemwould require pro-hibitiveamounts of computer time. Amathematical model for predictingemis-sion of pollutants due to incomplete combustionmust include the followingsubmodels: A two- or three-dimensional fluid dynamic code which accountsfor the effects of backmixing, buoyancy, and densityfluctuationsof bulkflow. Aturbulence model withcoefficients that depend on local Reynoldsnumber andwhich takes intoaccount themean densitygradients andfluctuations. The turbulencemodel under consideration requiresfully-developed turbulent flowwith Reynolds numbers larger than10,000. However, typical Reynolds numbers incontinuous flareflames are less than 20,000and the model will averageall thedynamic responses of aflame. Afinite,reactionratemechanismincludingamodel of soot forma-tion and burnout. An approximationwhichdescribes theeffects of turbulence on'mean reaction rates based on species and temperature fluctuations. A formulation of simultaneous gas and soot radiation. Amodel of radiation self-absorptionwhichconsiders the inter-mittencyof the turbulent flame.For the most part, versions of eachof the submodels are available,though integrating themfor solution to the problems of flares usingiter-ative POEmethods is difficult. Buildupof an overall flare model withthesame levelof sophisticationas the submodels mayyieldasystembeyond thehandlingcapacityof present-daycomputers. The POEmethods available forsolutionto turbulent jet flames (Hutchinson, et al, 1976; Janicka, 1979;and Pope, 1981), usuallyneglect or simplifythe chemistryand heat transfer.Solutions found by these complicatedmethods explaindata no better thanthose obtainedfrommuch simpler approaches.2-100-Solutions generated by POE are expensive and require agreat dealof interpretation. For instance, just one POE submodel developed by Dr. Wolf-gang Richter of EER requ1res.one-halfan hour of runtime on a CDC 7076 andcosts several thousand dollars. Output fromthese models must then be con-verted to common engineering parameters for use indesig