1. Prediction and Measurement of Flare Ignition Using the LES based C3d www.inl.govJ. D. Smith, Ph.D., Idaho National Laboratory A. Suo-Anttila, Ph.D., Systems Analyses and SolutionsS. Smith and N. Philpot, Zeeco, Inc. American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in Combustion Technology: Improving the Environment and Energy EfficiencySheraton, Maui, Hawaii September 26-29, 2010

2. Slide 2 OUTLINE Background and Introduction Flare Tests Model Setup and Methodology Simulation Results Low Flow Conditions High Flow Conditions Observations and ConclusionsMaui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 3. Slide 3 Elevated Multi-Tip Gas Flare Ignition Nominal Firing Rate = 350 Tons Per Hour (TPH) Max Firing Rate 1350 TPH Mostly Natural Gas (Mwt = 18) Experienced Pressure Wave during ignition Conducted Tests to quantify ignition phenomena: Microphones used to measure pressure wave High Speed Video used to capture flame during ignition Test results reported elsewhere (summarized below) Test video shows ignition behavior Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 4. Slide 4Test Layout Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 5. Slide 5 Flame during Ignition Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 6. Slide 6 Test Results:Sound Level (pressure wave) and Flame speed estimate Flame speed estimate from tests is 45 m/s(Test 1) and 50 m/s (Test 2) Maximum pressure generated by sphericalflame propagating at 50 m/s would be ~48mB (AIChE correlation*)* Center for Chemical Process Safety, Guidelines for Evaluating the Characteristics of Vapor CloudExplosions, Flash Fires and BLEVEs. AIChE (1994). Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 7. Slide 7Model Setup: General Comments Transient conservation equations with radiative heat transfer and combustion chemistry Considers soot formation and other multi-phase systems using Eulerian/Eulerian formulation Accurately assess different operation scenarios (wind, flow rate, fuel type, surroundings) Reasonable CPU time requirements on standard workstation Trade offs for Engineering Approach Sacrifice generality (large fires only) in favor of quick turnaround withquantitative accuracy Reaction rates and radiation heat transfer models apply only to large fires Models intended to make predictions good-enough for industrial use Model validation for each application to establish accuracy of resultsMaui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 8. Slide 8Combustion Model Variant of Said et al. (1997) turbulent flame model Relevant Species (model includes relevant reactions)F = Fuel Vapor (from evaporation or flare tip)O2 = OxygenPC = H20(v) + CO2C = Radiating Carbon SootIS = Non-radiating Intermediate Species Eddy dissipation effects and local equivalence ratio effects Reactions based on Arrhenius kineticsC and TA determined for all reactions Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 9. Slide 9 Reaction Rate Model Arrhenius rate model Consumption of primary reactant increases on reactants mass fraction fRi andtemperature T in volume Coefficients C and Activation Temperatures (TA) determined for all reactions Where:Ak = Pre-exponential FactorX1 = Natural Gas Mol FracX2 = O2 Mol Frac Ea = Activation TemperatureT = Local Gas Temperatureb, c, d = Global ExponentsMaui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 10. Slide 10 Approach (1) Laminar flame global mechanism used as starting point Used Activation temperature + mol frac exponents (based on reaction) Pre-exponential (Ak) factor adjusted to match turbulent combustion rxn rates Turbulent mixing effect on combustion included via LES Two coefficients adjust effect ( = turbulence intensity scale factor; = combustion species mixing time delay) Parameters adjusted to to match experimental results Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 11. Slide 11 Approach (2) Other model variations considered: Computational grid size & cell number Turbulence model (zero equation and one-equation LES) Nozzle structure (jet cone vs nozzle surface) Numerical upwind differencing Time to ignition Boundary Conditions Hydrostatic pressure on all external boundaries (adjusted to account for wind) Initial temperature and composition set to 300 K and air Other Assumptions Flare gas combustion approximated as described above Thermal radiation calculated w/ standard radiation models Wind conditions, flare gas inlet temperature and pressure, and radiation effects set to match measured value Flame emissivity = F (gas comp, soot fv, flame size/shape, temp) Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 12. Slide 12 Nozzle Approximation Two approaches considered for detailed nozzle structure: 1.Mass Sources on Nozzle Cone: Place source terms on cone surface and inject natural gas at correct velocity and mass rate as if resolved using fine cells Individual nozzle flow kept exactly identical (in absence of any flow mal-distribution) 2.Mass Inlet on Nozzle Surface Inject fluid through cells representing nozzles Total inlet adjusted for correct mass flow Individual rates varied (nozzle sizes varied due to overlap of square cells on circular tips) 3.More general mass source approach selectedMaui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 13. Slide 13 Computational Mesh o Flare dimensions approximated as 3.35m square, computational volume set as 20m long X 20m wide X 15m high (domain extended ~9m beyond flare edge) o Domain bottom set at top of elevated flare exit (reduce mesh size) o Domain separated into two regions Region 1: Near Tip Region just above nozzle and 7m square by 8m high Fixed horizontal cells with equal spacing (80 cells 0.0875m on a side) Vertical dimension slowly varied with 0.05m at nozzle face to 0.14m at top of region (90 cells) High resolution region had 576,000 cells Region 2: Buffer Region surrounding Near Tip Region Course, stretched cells to provide buffer between boundaries and near tip Both horizontal dimensions included 14 cells; vertical dimension included 12 cells Provided large distance from edge of domain (pressure boundary) and flame surface to prevent estimated pressure in igniting flame ballo During analysis, mesh refined several times to improve calculation results Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 14. Slide 14 Pressure Monitoring Locations Note: x and z are horizontal positions (x = 0 and z = 0 is flare center) and y is height above flare tip as shown in graphic)Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 15. Slide 15 Model Tuning (1) Over 60 CFD runs indicated pressure wave magnitude mostly dependent onignition time (combustion kinetics and turbulence had secondary effects) Typical pressure pulse of +30 to +40 mB wave followed by negative wave of -10to -20 mB Runs with ignition delay exhibited higher pressures waves Combustion parameters varied over significant range but had little effect on predicted peak pressure wave Ignition delay accomplished by: Natural gas jets turned on for 0.25 sec prior to igniting pilot After ignition, pilot flame grew and ignited flare gas at approximately 1 sec Resulting flame ball had significantly higher pressures than nearly all other cases considered Cases #3 and #41 had overpressures ~0.5 atm Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 16. Slide 16Predicted Pressure for All Cases at Low Flow Conditions Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 17. Slide 17 Model Tuning (2) Final chemical kinetics coefficients selected as providing best fit to ignitiontests: Ak = 5.0e16, Ta = 20098, b = 0.5, c = 1, and d = 1 Turbulence parameters selected: = 0.2; = 1e-5 Kinetics and turbulence parameters not highest values tested (i.e. fastestkinetics and most rapid mixing) Cases with higher values not always result in higher pressures since highvalues also leads to combustion in non-ideal mixtures Increasing turbulence scale improves mixing and suppresses natural fluidoscillations in turbulent jet (scale factor not allowed to exceed 2x recommendedvalue of 0.2) Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 18. Slide 18 Low Flow Results: Filtering Effect Case 18Case 3742 mB (0.62 psig)0.77 psig-10 mB (0.15 psig) Time history plot of local gas pressure for typical case @ 4m elevation above nozzle LHS figure has each point representing average of 4 time steps (slight filtering) Without filtering (RHS figure), isolated pressure peaks for single time steps (< 0.1 ms) predicted considered not representativeof experimental measurements (too fast for test equipment to accurately monitor) Filtering used to insure pressure histories representative of large regions and times more consistent with pressure historiesinferred from flame velocity measurements Pressure change (max min) reaches approximately 50 mB (or more) same as reported in flare tests Time between max and min pressure is on order of 16 ms (~60 Hz sound frequency)Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 19. Slide 19 Low Flow Results:Ignition Delay Effect Case 3 Case 41425 mB (6.25 psig) 125 mB (1.84 psig)8.16 psig 2.72 psig-60 mB (0.88 psig) -130 mB (1.91 psig) Pressure histories from two delayed ignition cases (Case 3 and Case 41) Highest Pressure observed on outer edge of growing fire ball Minimum pressure observed at center of growing fire ball after high pressure propagates outwardMaui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 20. Slide 20 Low Flow Results:Pressure Spikes from Ignition Delay Highest Pressure on outer edge of growing fire ball Minimum pressure at fire ball center after high pressureregion propagates outward Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 21. Slide 21 Low Flow Results:Horizontal temperature contour through flame ball at 4mBeginning of IgnitionMid-Point of Ignition Near End of Ignition Spatial distance between tick marks on plots is 1m; temporal distance between plots is 30 ms Dividing flame propagation distance by time between frames yields flame velocity of 33 m/s Experimental flame propagation velocity ~50 m/s (examining video data indicated they failed to subtractinitial ball diameter). Correcting test results yields actual growth rate of 40 44 m/s Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 22. Slide 22 Low Flow Results:Flame ball growth for normal and delayed ignition Normal IgnitionDelayed Ignition Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 23. Slide 23 Low Flow Results:Normal Ignition Video Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 24. Slide 24 Low Flow Results:Delayed Ignition Video Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 25. Slide 25 High Flow Results:Ignition Delay Effect High Rate wo/ Ignition Delay High Rate w/ Ignition Delay 1700 mB (24.98 psig)160 mB (2.35 psig) 3.53 psig -80 mB (1.18 psig) Pressure history from non-delayed (RHS) and delayed (LHS) ignition pressure wave (4m above flare tip) Higher flow wo/ ignition delay caused slightly higher pressure wave (3.53 vs 0.77 psig) With Ignition delay, pressure builds until calculation becomes unstable (detonation)Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 26. Slide 26 High Flow Results:Normal Ignition Video Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 27. Slide 27High Flow Results: Delayed Ignition Video (only two frames at sampling rate) Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 28. Slide 28Conclusions Natural Gas Flare Gas Fired through multi-burner tip: C3d flare model based on LES mixing model Combustion model used EBU type reactions (includes soot) 2-zone computational mesh (adjusted to optimize grid) Final mesh size ~1.2MM cells Simulated low flow (200-350 TPH) and high flow (1350 TPH) conditions Compared results to test results Pressure wave estimated by AIChE correlation + flame speed estimated from high speed video (pressure measurements via microphone not sensitive enough) Predictions compared well to data for flame speed and pressure wave from 12 tests (2 tip sizes, 3 operating pressures, 2 radiation sample locations) Estimated Pressure wave Low flow, no ignition delay < 0.75 psig, flame speed ~33 m/s (measured 40 m/s) Low flow, ignition delay ~ 8 psig possible! High flow, no ignition delay > 3.5 psig High flow, ignition delay resulted in explosion! Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 29. Slide 29 Backup Slides Maui, Hawaii September 27 - 29, American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 30. Slide 30Radiation Inside Large Fires High soot volume fractions make large fires non-transparent(optically thick) which causes flame to radiate as a cloud(radiatively diffuse) Fire volume defined where soot volume fraction (fi) greater thanminimum volume fraction (fsoot > fmin) Flame edge (fflameedge) where soot volume fraction = 0.05 ppm Calculated flame surfaces from 3 time steps from validation against test Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 31. Slide 31Radiation Outside of Large Fires When fsoot < fflameedge then outside flame (participating mediumconsidered) View factors from fire to surrounding surfaces calculated at eachtime step (includes attenuation by gas and soot media forflames) Re-radiation from surroundings also calculated at each time Fire considered black body radiator (firesurface = 1) Radiation from flame to surroundings assumes Tsurround =constant Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 32. Slide 32Diffuse Radiation Within Fire Calculated indirectly using a Rossland effective thermalconductivity 316T kR = >> k Air 3 R = Stefan-Boltzman Constant T = local temperature R= local extinction coefficient. Dependent on local species concentrations Radiation transport model: Predicts radiant flux on external (and internal) surfaces Provides source/sinks terms to overall energy equation Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 33. Slide 33 Reactions Involving Fuel Incomplete Fuel Combustion (soot producing)1 kg F + (2.87-2.6S1) kg O2 S1 kg C + (3.87-3.6S1) kg PC + (50-32S1) MJ Combustion Soot Mass Parameter, S1 = 0.005 Endothermic Fuel Pyrolysis (soot producing)1 kg F + 0.3 MJ S2 kg C + (1-S2) kg IS Cracking Parameter, S2 = 0.15 Maui, HawaiiAmerican Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010 Combustion Technology: Improving the Environment and Energy Efficiency Sheraton 34. Slide 34 Reactions Not Involving Fuel Soot Combustion 1 kg C + 2.6 kg O2 3.6 kg CO2 + 32 MJ Combustion of Intermediate Species Coefficients chosen so that complete combustion of C and IS produce samespecies and thermal energy as direct combustion of fuel Maui, Hawaii American Flame Research Committees - International Pacific Rim Combustion Symposium Advances in September 27 - 29, 2010Combustion Technology: Improving the Environment and Energy Efficiency Sheraton