Transcription

A Methodology to Relate Black CarbonParticle Number and Mass Emissions fromVarious Combustion SourcesRoger Teoh1, Marc Stettler1, Arnab Majumdar1 & Ulrich Schumann21Centre for Transport Studies, Department of Civil and Environmental Engineering, Imperial CollegeLondon, London, UK2Deutsches Zentrum fΓΌr Luft- und Raumfahrt, Institute of Atmospheric Physics, 82234Oberpfaffenhofen, Germany2018 Cambridge Particle Meeting15th June 20181

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Further Work2

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Further Work3

Why is BC PN Emissions Important? Health Effects: Ultrafine particles have a higher probability of being deposited into the respiratorysystem, and translocated towards the circulatory system and internal organs. Climate Effects: During cruise, Black Carbon (BC) Particle Number (PN) emissions from aircraft enginesact as a condensation nuclei for contrail formation. No. of contrail ice particles No. of aircraft BC particle emissions per kg-fuel burnt (EIn in kg-1) Different young contrail characteristics influenced by BC EIn.𝜏 Optical depth Existing BC EIn models for aviation emission assumes that BC particlemorphologies remain constant irrespective of engine thrust settings. BC mass measurements and estimates remain more commonlyavailable than the number metric.Source: [1]4

Research Objectives1) Develop a new model to estimate BC particle number emissions from mass based onthe theory of fractal aggregates.2) Validate the new model using BC measurements from three different emission sources.3) Perform an uncertainty and sensitivity analysis to understand the accuracy anduncertainty bounds of the newly developed model.5

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Further Work6

Development of a New BC PN Predictive Model Mass of one BC aggregate (m) is the summation of primaryparticle masses:π‘š 𝑛pp 𝜌0whereπœ‹π‘‘pp 36𝑛pp Number of primary particles in an aggregate𝜌0 BC material density (1770 kg/m3)𝑑pp Primary particle diameter The total mass of aggregates (M) is calculated using theintegrated product of the aggregate mass and numberweighted distribution: 𝑀 0whereπ‘š 𝑑m 𝑛 𝑑m d𝑙𝑛𝑑m𝑛 𝑑m 𝑁 𝑝(π‘‘π‘š )Source: [11]7

π‘š 𝑛pp 𝜌0πœ‹π‘‘pp 36Development of a New BC PN Predictive Model In the free-molecular regime, the number of primary particles in an aggregate (npp) [6]:𝑑m 2𝐷) α𝑑pp𝑛pp π‘˜a (where 𝑑m 𝐷) fm𝑑pp𝑛pp (or𝑑m Aggregate mobility diameterπ‘˜a Scaling pre-factor𝐷α Projected area exponentDfm Mass-mobility exponentEggersdorfer et al. (2012) [7] suggested universal values of π‘˜a 0.998 and 𝐷𝛼 1.069 foraggregates formed of polydisperse primary particles, irrespective of the state of sintering. The Knudsen Number (Kn) is a dimensionlessnumber equal to the ratio of the mean freepath (πœ†) to the particle radius:𝐾𝑛 Free-molecular regime7% F/F00Continuum regime65% F/F00100% F/F002πœ†π‘‘ Free-molecular regime: Kn 1 Continuum regime:Kn 1 Transition regime:0.1 Kn 108

π‘š 𝑛pp 𝜌0πœ‹π‘‘pp 36Development of a New BC PN Predictive Model Relationship between primary particle (dpp) and aggregate mobility diameter (dm) [9]:𝑑pp π‘˜TEM 𝑑m 𝐷TEMwhere prefactor-exponent coefficient pairs π‘˜TEM & 𝐷TEM are fitted with Transmission Electron Microscopy (TEM)GDI engine:Aviation gas turbine:π‘˜TEM 2.616 10 6π‘˜TEM 1.621 10 5𝐷TEM 0.30𝐷TEM 0.39HPDI engine:Inverted burner:π‘˜TEM 2.644 10 6π‘˜TEM 2.465 10 6𝐷TEM 0.29𝐷TEM 0.29dm (nm)dm (nm)Source: [9]9

Development of a New BC PN Predictive Model The total mass of aggregates (M) for a given particle size distribution:𝑑m 2𝐷) α𝑑pp𝑛pp π‘˜a (π‘š 𝑛pp 𝜌0𝑑pp π‘˜TEM 𝑑m 𝐷TEMπœ‹π‘‘pp 36𝑛 𝑑m 𝑁 𝑝(π‘‘π‘š ) 𝑀 0π‘š 𝑑m 𝑛 𝑑m d𝑙𝑛𝑑mπœ‹π‘€ π‘π‘˜a 𝜌0 ( )(π‘˜TEM )3 2𝐷𝛼6 0𝑑m πœ‘ p 𝑑m d𝑙𝑛𝑑mwhere πœ‘ 3𝐷TEM 1 𝐷TEM 2𝐷𝛼Source: [11]10

Development of a new BC PN Predictive Model The remaining integral is the Ο†th moment of a log-normal distribution:πœ‹π‘€ π‘π‘˜a 𝜌0 ( ) π‘˜TEM6whereGMDGSD Rearrange for N:𝑁 πœ‹π‘˜a 𝜌0 ( ) π‘˜TEM63 2π·π›Όπœ‘2 ln GSD 2GMD exp()2πœ‘Geometric Mean DiameterGeometric Standard Deviation𝑀3 2𝐷𝛼 GMDπœ‘ exp(πœ‘2 lnGSD 2)2New N or EIn PredictiveModel, called the FractalAggregates (FA) Model. Advantages: New FA model relates BC mass, number and Particle Size Distribution (PSD) in one equation. Captures the change in particle morphology for different combustion conditions11

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Further Work12

Model Validation – CIDI Engine & Inverted Burner𝑁 π‘€πœ‹π’Œπš 𝜌0 ( )6πœ‘2 ln GSD 2π‘˜TEM 3 2π‘«πœΆ GMDπœ‘ exp()2(a) Experimentally fitted π’Œπ’‚ and π‘«πœΆ valuesfor each operating mode [14]CIDI Engine Data Source:Inverted Burner Data Source:where πœ‘ 3𝐷𝑇𝐸𝑀 1 𝐷𝑇𝐸𝑀 2π‘«πœΆ[12][13](b) Constant π’Œπ’‚ 𝟎. πŸ—πŸ—πŸ– and π‘«πœΆ 𝟏. πŸŽπŸ”πŸ— forall operating mode [7]13

Model Validation – Aircraft Gas Turbine EnginesConstant π’Œπ’‚ 𝟎. πŸ—πŸ—πŸ– and π‘«πœΆ 𝟏. πŸŽπŸ”πŸ— for all operating mode [7]EIn EImπœ‹πœ‘2 ln GSD 2π’Œπš 𝜌0 ( ) π‘˜TEM 3 2π‘«πœΆ GMDπœ‘ exp()6where πœ‘ 3𝐷𝑇𝐸𝑀 1 𝐷𝑇𝐸𝑀 2π‘«πœΆ2(a) GroundValidation(b) CruiseValidation When ka and DΞ± values of 0.998 and 1.069 [7] are applied to aircraft datasets, on average, we obtainnegative R2 and 100% NMB valuesSource:SAMPLE III.2[10]Source:NASA ACCESS[15]14

Model Validation – Aircraft Gas Turbine EnginesConstant π’Œπ’‚ 𝟎. πŸ—πŸ—πŸ– and π‘«πœΆ 𝟏. πŸŽπŸ”πŸ— for all operating mode [7]EIn EImπœ‹πœ‘2 ln GSD 2π’Œπš 𝜌0 ( ) π‘˜TEM 3 2π‘«πœΆ GMDπœ‘ exp()6where πœ‘ 3𝐷𝑇𝐸𝑀 1 𝐷𝑇𝐸𝑀 2π‘«πœΆ2(a) GroundValidation(b) CruiseValidationdm (nm)dm (nm)dm (nm) When ka and DΞ± values of 0.998 and 1.069 are applied to aircraft datasets, on average, we obtainnegative R2 and 100% NMB valuesSource:SAMPLE III.2[10]Source:NASA ACCESS[14]15

Model Validation – Aircraft Gas Turbine EnginesEIn EImπœ‹6𝟏 𝜌0 ( ) π‘˜TEM Recall: 𝒏𝒑𝒑 π’Œπ’‚ (πœ‘2 ln GSD 23 π‘«π’‡π’ŽGMDπœ‘ exp()2π’…π’Ž πŸπ‘«) πœΆπ’…π’‘π’‘or 𝒏𝒑𝒑 (π’…π’Ž π‘«π’‡π’Žπ’…π’‘π’‘)where πœ‘ 3𝐷𝑇𝐸𝑀 1 𝐷𝑇𝐸𝑀 π‘«πŸπ¦12. Hence, we assume that π‘˜π‘Ž 1 and 𝐷𝛼 𝐷fm [7](b) CruiseValidation(a) GroundValidationSource:SAMPLE III.2[10]Source:NASA ACCESS[15]16

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Future Work17

Results: Uncertainty Analysis Uncertainty analysis is performed using the MonteCarlo 1000-member ensembles [16]. The uncertainty of the estimated EIn are asymmetricallydistributed (-37%, 55%) at 1.96Οƒ. The asymmetrical distribution is due to the non-linearityof the FA model. An uncertainty analysis on the estimated EIn was notconducted in previous aviation PN methodologies.18

Results: Sensitivity Analysis A variance-based global sensitivity analysis identifiedthat the uncertainties in GSD contribute to the largestsensitivity in the FA model output.Sensitivity Analysis(Measured Input Parameters) A prioritisation can be recommended for futureresearch to measure certain variables (such as M, Dfm,GMD and GSD) more accurately to reduce theuncertainty bounds of the FA model outputs. Given that ka contributes to the lowest sensitivity to theestimated EIn, the assumption of π‘˜a 1 across allengine types & F/F00 would not significantly affect theFA model outputs.19

Outline1.Introduction & Motivation2.Derivation of a new BC N or EIn Predictive Model3.Model Validation4.Uncertainty & Sensitivity Analysis5.Conclusions & Further Work20

Conclusions A new methodology to relate BC ParticleNumber and Mass emissions is developedbased on the theory of fractal aggregates.𝑁 π‘€πœ‹π‘˜a 𝜌0 ( 6 ) π‘˜TEM3 2𝐷𝛼 GMDπœ‘ exp(πœ‘2 lnGSD 2)2 The new FA Model is validated with threedifferent emission sources: An internal combustion engine (CIDI)An inverted burnerTwo aircraft gas turbine engines An uncertainty analysis estimates N or EIn tohave an asymmetrical uncertainty bound(-36%, 54%) at 1.96Οƒ. A sensitivity analysis shows that GSD is themost critical input parameter, followed by the M,Dfm and GMD.21

Future WorkFA Model Application to Aviation Emissions:Is there a net climate benefit in diverting flights to avoid contrail formation?Aircraft Activity Dataset:CARATS Open Data, JapanEmissions:CO2, BC EIn (FA Model) & EIm, Ambient atmospheric conditions:Contrail Model:ECMWF DatasetCoCiPClimateImpactsAir TrafficManagement22

Thank you, [email protected] Postgraduate,Centre for Transport Studies,Department of Civil and Environmental EngineeringImperial College LondonAcknowledgements:This PhD is funded by The Lloyds Register Foundation, and the Skempton Scholarship from the Department ofCivil and Environmental Engineering, Imperial College London.23

References[1] KÀrcher, B. (2016) The importance of contrail ice formation for mitigating the climate impact of aviation. Journal of Geophysical Research:Atmospheres. 121 (7), 3497-3505.[2] Petzold, A., Dâpelheuer, A., Brock, C. & Schrâder, F. (1999) In situ observations and model calculations of black carbon emission by aircraft at cruisealtitude. Journal of Geophysical Research: Atmospheres. 104 (D18), 22171-22181.[3] Dâpelheuer, A. (2002) Anwendungsorientierte Verfahren Zur Bestimmung Von CO, HC Und Ruß Aus Luftfahrttriebwerken.[4] Barrett, S., Prather, M., Penner, J., Selkirk, H., Balasubramanian, S., Dâpelheuer, A., Fleming, G., Gupta, M., Halthore, R. & Hileman, J. (2010)Guidance on the use of AEDT gridded aircraft emissions in atmospheric models. US Federal Aviation Administration Office of Environment and Energy.[5] Lee, D., Pitari, G., Grewe, V., Gierens, K., Penner, J., Petzold, A., Prather, M., Schumann, U., Bais, A. & Berntsen, T. (2010) Transport impacts onatmosphere and climate: Aviation. Atmospheric Environment. 44 (37), 4678-4734.[6] Sorensen, C.M., 2011. The mobility of fractal aggregates: a review. Aerosol Science and Technology, 45(7), pp.765-779[7] Eggersdorfer, M.L., Kadau, D., Herrmann, H.J. and Pratsinis, S.E., 2012. Aggregate morphology evolution by sintering: number and diameter of primaryparticles. Journal of aerosol science, 46, pp.7-19.[8] Liati, A., Brem, B.T., Durdina, L., Vögtli, M., Arroyo Rojas Dasilva, Y., Dimopoulos Eggenschwiler, P. and Wang, J., 2014. Electron microscopic study ofsoot particulate matter emissions from aircraft turbine engines. Environmental science & technology, 48(18), pp.10975-10983.[9] Dastanpour, R. & Rogak, S. N. (2014) Observations of a correlation between primary particle and aggregate size for soot particles. Aerosol Scienceand Technology. 48 (10), 1043-1049.[10] Boies, A. M., Stettler, M. E., Swanson, J. J., Johnson, T. J., Olfert, J. S., Johnson, M., Eggersdorfer, M. L., Rindlisbacher, T., Wang, J. & Thomson, K.(2015) Particle emission characteristics of a gas turbine with a double annular combustor. Aerosol Science and Technology. 49 (9), 842-855.[11] Lobo, P., Hagen, D.E., Whitefield, P.D. and Raper, D., 2015. PM emissions measurements of in-service commercial aircraft engines during the DeltaAtlanta Hartsfield Study. Atmospheric Environment, 104, pp.237-245.24

References[12] Graves, B., Olfert, J., Patychuk, B., Dastanpour, R. and Rogak, S., 2015. Characterization of particulate matter morphology and volatility from acompression-ignition natural-gas direct-injection engine. Aerosol Science and Technology, 49(8), pp.589-598.[13] Dastanpour, R., Momenimovahed, A., Thomson, K., Olfert, J. and Rogak, S., 2017. Variation of the optical properties of soot as a function of particlemass. Carbon, 124, pp.201-211.[14] Dastanpour, R., Rogak, S.N., Graves, B., Olfert, J., Eggersdorfer, M.L. and Boies, A.M., 2016. Improved sizing of soot primary particles using massmobility measurements. Aerosol Science and Technology, 50(2), pp.101-109.[15] Moore, R. H., Thornhill, K. L., Weinzierl, B., Sauer, D., D’Ascoli, E., Kim, J., Lichtenstern, M., Scheibe, M., Beaton, B. & Beyersdorf, A. J. (2017)Biofuel blending reduces particle emissions from aircraft engines at cruise conditions. Nature. 543 (7645), 411-415.[16] Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M. & Tarantola, S. (2008) Global sensitivity analysis: the primer. ,John Wiley & Sons[17] Schumann, U., 2012. A contrail cirrus prediction model. Geoscientific Model Development, 5, pp.543-580.[18] Abegglen, M., Durdina, L., Brem, B. T., Wang, J., Rindlisbacher, T., Corbin, J. C., Lohmann, U. & Sierau, B. (2015) Effective density and mass–mobility exponents of particulate matter in aircraft turbine exhaust: Dependence on engine thrust and particle size. Journal of Aerosol Science. 88 135-147.25

Global WarmingSource: /shutterstock 91110830.jpg26

Existing BC EIn Models for Aviation Emissions1)2)3)4)Average BC Particle Mass [2] Total BC mass divided by an average mass ofeach BC particle mBC EIn (3.2 10 17)BC EIEIn/EIm Ratio with Altitude Variations [3] 5x1015 to 1.6x1016 particles per g(BC) Used in the Aero2K Global Aviation EmissionsInventoriesAssumed Particle Diameter [4]EImπœ‹93( )πœŒπ‘π‘‰ GMD exp(62 EIn Assume log-normal distribution, GMD andGSD fixed at 38nm and 1.6 respectively.lnGSD 2 )BC EIn Range [1] BC EIn 1014 to 1015 per kg fuel burnedKEY LIMITATION:Existing BC EIn models do not include adependence on the change in BC aggregatemorphologies vs engine thrust settings27

Knudsen Number, KnKnudsen Number, 𝐾𝑛 2πœ†π‘‘ Mean free path, πœ† Average distance travelled by an aggregate between successive collisionswith gas molecules. The mean free path of at a given pressure (P1) can be estimated using standard atmosphericconditions (P0, πœ†0 ) as a reference:𝑃𝑃1πœ†1 πœ†0 0 ,whereP0 1 atm & πœ†0 0.066 πœ‡π‘š For the emissions from an internal combustion engine and aircraft gas turbine, the pressure inthe combustor is used for 𝑃1 For inverted burner emissions, we assume that P1 1 atm28

Knudsen Number, KnFree-molecular regime7% F/F00 Free-molecular regime: Kn 1 Continuum regime:Kn 1 Transition regime:0.1 Kn 10Continuum regime65% F/F00Sources: [8], [18]100% F/F0029

How is ka and DΞ± Experimentally Fitted?Dastanpour et al. (2016) [14]METHOD 1: Estimation of ka and DΞ± using Optimisation Combining equations 𝑛𝑝𝑝 π‘˜π‘Ž (π‘‘π‘š 2𝐷𝛼𝑑𝑝𝑝)𝑑𝑝𝑝 πœ‹and π‘š π‘›π‘π‘π‘˜π‘Ž πœ‹πœŒ0π‘‘π‘š6π‘š62𝐷𝛼𝑑𝑝𝑝 3 𝜌0 to obtain:12𝐷𝛼 3Optimum ka and DΞ± values are calculated using regression to minimise the difference between theTEM determined dpp and the above equation.METHOD 2: Estimation of ka and DΞ± using Experimental Measurements π‘‘π‘š 2𝐷) 𝛼𝑑𝑝𝑝Combining equations 𝑛𝑝𝑝 π‘˜π‘Ž (In and 𝑛𝑝𝑝 π‘šπ‘šπ‘π‘to obtain:π‘šπ‘‘π‘š 2𝐷𝛼 ln ln(π‘˜π‘Ž )π‘šπ‘π‘π‘‘π‘π‘Mass-mobility data (obtained from CPMA and DMA), and TEM-obtained dpp and mpp were used inthe above equation to obtain ka and DΞ± values.30

Uncertainty & Sensitivity Analysis𝐸𝐼𝑛 VariableBC EIm (LII)𝜌0kaDfmkTEMDTEMGMDGSDFixed F/F000.40.40.40.40.40.40.40.4πΈπΌπ‘šπœ‹π‘˜a 𝜌0 ( ) π‘˜TEM6Mean (Β΅)2.7 mg/kg1770 kg/m312.761.621x10-50.3918.49 nm1.733 2𝐷𝛼 GMDπœ‘ exp(πœ‘2 lnGSD 2)2Std Dev *ΞΌ- Uncertainty Distribution for all Parameters: Normal31

Model Application to Aviation EmissionsCan we apply the new FA model to estimate BC EIn for globalcivil aviation, or at an individual flight level?EIn EImπœ‘2 ln GSD 2πœ‹πœŒ0 ( ) π‘˜TEM 3 𝐷fm GMDπœ‘ exp()62where πœ‘ 3𝐷TEM (1 𝐷TEM )𝐷fmRequirements: Estimate different input variables (BC EIm, GMD, GSD and Dfm) versus engine thrustsettings (F/F00).32

Model Inputs – (1) Existing BC EIm ModelsEIn π„πˆπ¦πœ‹πœŒ0 ( ) 1.621 10 563 𝐷fm GMDπœ‘ exp(πœ‘(a) Ground Conditions2 lnGSD 2)2(b) Cruise ConditionsUncertainty(EIm): 50%Source:FOA3 [16]FOX [17]ImFOX [18]33

Model Inputs – (2) GMD & (3) GSDEIn EImπœ‹πœŒ0 ( ) 1.621 10 56(a) BC GMD vs. Thrust Settings (F/F00)Uncertainty(GMD): 25%3 𝐷fm π†πŒπƒπœ‘ exp(πœ‘2 ln𝐆𝐒𝐃 2)2(b) BC GSD vs. Thrust Settings (F/F00)Uncertainty(GSD): 15%34

Model Input Parameters – (4) DfmEIn EImπœ‘ 2 ln GSD 2πœ‹πœŒ0 ( ) 1.621 10 5 3 π‘«πŸπ¦ GMDπœ‘ exp()62whereπœ‘ 1.17 0.61π‘«πŸπ¦πΉπ·π‘“π‘š 2.04, for 0.03 𝐹 0.200πΉπ·π‘“π‘š 2.35, for 0.2 𝐹 0.5000πΉπ·π‘“π‘š 2.64, for 0.5 𝐹 100*Dfm values are for Singular AnnularCombustor (SAC) turbofan engines onlyUncertainty (Dfm): 27%Source: [19]35

FA Model Validation – Estimated Inputs Cruise BC EIn and EIm measurementswere available from the SULFURexperimental campaigns. [20] [21] No GMD, GSD and Dfm measurementswere available. Estimated inputs for GMD, GSD andDfm versus F/F00 (specified in previousslides) were used. Validation results justify the use of theGMD, GSD and Dfm predictive relationsas model inputs to the FA model.36

A Methodology to Relate Black Carbon Particle Number and Mass Emissions from Various Combustion Sources Roger Teoh 1, Marc Stettler , Arnab Majumdar1 & Ulrich Schumann2 1 Centre for Transport Studies, Department of Civil and Environmental Engineering, Imperial College London, London, UK 2