Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Investigations of fuel effects on turbulent premixed jet flames
(USC Thesis Other)
Investigations of fuel effects on turbulent premixed jet flames
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INVESTIGATIONS OF FUEL EFFECTS ON TURBULENT PREMIXED JET FLAMES BY JENNIFER SMOLKE DECEMBER 2017 A DISSERTATION PRESENTED TO THE FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA IN PARTIAL FULLFILLMENT OF THE REQUIREMENTS FOR THE DEGREE DOCTOR OF PHILOSOPHY (MECHANICAL ENGINEERING) Acknowledgements This dissertation was the result of countless conversations, advice, and collaborations with many people in my department and otherwise. First, I would like to thank Dr. Egolfopoulos for encour- aging me to think outside of the box and for inspiring this project. I would also like to acknowledge Dr. Fincham, Dr. Spedding, and Dr. Sellappan for their help and advice on Particle Image Ve- locimetry post-processing. I would like to thank Dr. Ronney who was instrumental in helping me to understand some of the basics of turbulent combustion and scaling. I would like to acknolwedge Dr. Francesco Carbone who initally worked with me to build the experimental apparatus from the ground up and taught me everything I know about how to be a good experimental researcher. His advice and ideas have been a large inspiration for the investiga- tions conducted. I would also like to thank Laurel Paxton for her collaboration on this project and for the many hours of discussion on results and for being a reliable friend in the program. I would also like to acknowledge Dr. Hai Wang at Stanford for the collaborative effort and inspiration for then-dodecane decomposition study. I would like to thank Dr. Blanquart and Dr. LaPointe at Cal- Tech as well for their insight and collaborative effort in LES investigations of our burner. I would also like to thank Vyaas for teaching me OpenFOAM, DJ, Jagan, and Roe for their leadership in the laboratory, and the rest of my labmates for discussions in and out of the lab. I would also like to thank Silvana and Kim who placed countless orders for me in order to turn equipment into a working experimental apparatus. I would like to thank Valerie, Sam, Irice, and Melissa who were always willing to give advice and who’s friendly faces made the USC environment much better. I would also like to acknowledge funding from the NSF fellowship for the past three years which allowed me to focus full time on this project. This work was also thanks to the sponsorship by Dr. Chiping Li at AFOSR which provided the funds to build the PIV system, burner, and high speed imaging. Finally, this dissertation is dedicated to my family who have loved and encouraged me every step of the way and put up with my highs and lows. I would also like to thank my friends Alisha, Kiera and Kat for encouraging me to maintain a good balance between emotional health and work. Thanks also to Lisa, Amy, Rachel, Seema, Tiffany, Jennifer, for being faithful friends throughout this process. A big thanks as well to my mentors Mike Purser and Dr. Karagozian for encouraging me to pursue the Ph.D. path. Proverbs 3:5-6 ii Contents Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Combustion Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Laminar Premixed Flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.3.1 Flame Stretch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3.2 Fuel Structure and Breakdown in Laminar Flames . . . . . . . . . . . . . 6 1.3.3 Molecular Transport and Heavy Hydrocarbons . . . . . . . . . . . . . . . 7 1.3.4 Counterflow Flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.4 Turbulent Premixed Combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.4.1 Basics of Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.4.2 Turbulence in a Free Shear Flow . . . . . . . . . . . . . . . . . . . . . . . 11 1.4.3 Regimes of Premixed Turbulent Combustion . . . . . . . . . . . . . . . . 12 1.4.4 Turbulent Burning Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.4.5 Fuel Effects on Turbulent Premixed Combustion . . . . . . . . . . . . . . 15 1.4.6 Objectives and Structure of Thesis . . . . . . . . . . . . . . . . . . . . . . 17 2 Experimental Approach 18 2.1 Burner Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.2 CH* Chemiluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.3 Particle Image Velocimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.3.1 Types of Particles and Particle Generation . . . . . . . . . . . . . . . . . . 24 2.3.2 Particle Image Displacement Errors andDt Optimization . . . . . . . . . . 25 2.3.3 Ignited Flame and Index of Refraction . . . . . . . . . . . . . . . . . . . . 26 2.3.4 Post-Processing Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.3.5 Uncertainty Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.4 Temperature measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 iii 3 Numerical Approach 28 3.1 PREMIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2 Senkin and Opposed Jet Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.3 Large Eddy Simulations (LES) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.4 Turbulent Jet Flame Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4 Effects ofn-Dodecane Decomposition on its Fundamental Flame Properties 31 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.2 Computational Setup and Rationale . . . . . . . . . . . . . . . . . . . . . . . . . 32 4.3 Pyrolysis ofn-C 12 H 26 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.4 Propagation of mixtures or air with partially decomposedn-C 12 H 26 . . . . . . . . 38 4.5 Extinction of crackedn-C 12 H 26 /air flames . . . . . . . . . . . . . . . . . . . . . . 40 4.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 5 Heat Loss and Fuel Effects on Global Characteristics of Piloted Premixed Jet Burner 45 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 5.2 Experimental and Numerical Approach . . . . . . . . . . . . . . . . . . . . . . . 47 5.2.1 CH* Chemiluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 5.2.2 Particle Image Velocimetry (PIV) . . . . . . . . . . . . . . . . . . . . . . 49 5.2.3 Temperature measurements . . . . . . . . . . . . . . . . . . . . . . . . . 49 5.3 Experimental/numerical validation . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.3.1 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.3.2 Jet development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.4 Global fuel effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 5.4.1 Global Flame Shape and Centerline Luminosity Profiles . . . . . . . . . . 52 5.4.2 Adiabatic Flame Height Study . . . . . . . . . . . . . . . . . . . . . . . . 54 5.4.3 ExperimentalT coflow effects onH fl . . . . . . . . . . . . . . . . . . . . . . 57 5.4.4 Experimental Re jet effects onH fl . . . . . . . . . . . . . . . . . . . . . . . 58 5.4.5 Experimental and numerical comparison ofH fl . . . . . . . . . . . . . . . 59 5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 6 Experimental Estimation of Turbulent Flame Speed and Characterization of Global and Local Observables for Lean Jet Fuel/Air Mixtures at High Ka 61 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 6.2 Experimental Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 6.2.1 Burner Operating Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 63 6.2.2 Chemiluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 iv 6.2.3 Temperature Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 65 6.3 Experimental/numerical validation . . . . . . . . . . . . . . . . . . . . . . . . . . 65 6.3.1 Laser-Based Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 6.4 Global Flame Observables: Results and Discussion . . . . . . . . . . . . . . . . . 71 6.4.1 Fuel Effects on Global Flame Shape . . . . . . . . . . . . . . . . . . . . . 71 6.4.2 Adiabatic Flame Height Study . . . . . . . . . . . . . . . . . . . . . . . . 73 6.4.3 Mie Scattering Results and Discussion . . . . . . . . . . . . . . . . . . . . 75 6.5 Temperature Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 6.6 PIV Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 6.6.1 Stitched PIV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 6.6.2 Local Turbulent Flame Speed Measurements . . . . . . . . . . . . . . . . 83 6.6.3 Conditional Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 6.6.4 Nested PIV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 6.7 OpenFOAM Turbulent Jet Flame Case Study . . . . . . . . . . . . . . . . . . . . 91 6.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 7 Conclusions and Recommendations 95 7.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 7.2 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 96 References 97 v List of Tables 1.1 Lumped Mechanism for High-Temperaturen-Dodecane Oxidation [4] . . . . . . . 7 4.1 Species and their mass fractions of during the initial stage ofn-dodecane oxidation in air over a range of time or %n-dodecane decompossed. The computation was made for = 1.0 and several initial temperatureT p;0 . . . . . . . . . . . . . . . . . 38 4.2 Unburned mixture conditions used in flame simulations. . . . . . . . . . . . . . . 39 5.1 Inlet Conditions and Specifications of Parametric Heated PPJB Investigation . . . . 47 5.2 Conditions of the experiments and simulations performed. Re t is the turbulent Reynolds number, Ka is the Karlovitz number, and u 0 is the peak turbulent in- tensity at x=D=15. The integral length scale L int is evaluated in the shear layer at x=D = 15. All values are calculated based on the kinematic viscosity of the unburnt mixture but the turbulent properties of the flow, i.e. L int and u 0 used in Re t are measured with the flame ignited. S L and f , were calculated using PRE- MIX [46]. f was calculated for all fuels, but since less than 7% variation in value exists between the fuels only the value for methane was used in the table. . . . . . 48 6.1 Measured and calculated scaling parameters for investigated PPJB conditions. In- tegral length andU 0 were measured at a distance ofx=D = 15 in the shear layer of the flame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 6.2 Resolution and Window Size of Nested PIV camera setup . . . . . . . . . . . . . . 69 6.3 Condition matrix for flame heights of jet fuels . . . . . . . . . . . . . . . . . . . . 71 vi List of Figures 1.1 Worldwide Electricity Generation [1] . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Typical Turbojet Engine Structure and corresponding Pressure and Temperature Traces [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Combustor Cutaway and Flow Path [2] . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Regimes of Premixed Turbulent Combustion . . . . . . . . . . . . . . . . . . . . . 13 2.1 Piloted premixed jet burner (PPJB) cutaway. All measurements are in mm. . . . . . 20 2.2 In-House Vaporization Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.3 Computed correlations of key flame markers against the heat release rate obtained by laminar flame modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.4 Graphic showingH fl determination from the CH* Luminosity Profiles . . . . . . . 23 4.1 Structures computed for atmospheric-pressure laminar premixed stoichiometricn- dodecane-air flames of unburned gas temperature of 403 K. . . . . . . . . . . . . . 33 4.2 Time history computed for the thermal decomposition of 1.12% (mol)n-dodecane in air at 1 atm and initial temperature of 1400 K under adiabatic and isochoric conditions, plotted in logarithmic (left panel) and linear, zoomed-in (right panel) time scales. The horizontal lines on the right side of the left panel indicate the equilibrium composition of gas-phase species. . . . . . . . . . . . . . . . . . . . 34 4.3 An illustrative diagram of the computational experiments performed. The dashed line encloses an overall adiabatic reactor comprised of four processes: 1) the fuel- air mixtureX 0 at an initial temperatureT 0 = 403 K and pressure of 1 atm is heated to the pyrolytic reactor temperatureT p;0 , 2) the mixture undergoes adiabatic, iso- baric thermal decomposition for a residence time of to obtain a partially reacted mixture X p; , 3) cooling the partially reacted mixture to reach T 0 0 from T p;1 (to ensure identical adiabatic flame temperature T ad as the natural laminar premixed flame) after 4) an adiabatic laminar premixed flame. . . . . . . . . . . . . . . . . 35 4.4 Temperature (top), and mass fraction-time history (bottom) computed for adiabatic and isobatic oxidation of n-dodecane-air mixtures at three equivalence ratios () with initial temperature of a)T p;0 = 1100 K, b)T p;0 = 1200 K, and c)T p;0 = 1300 K, all at 1 bar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 vii 4.5 (a-c) Mass percentages of key products computed forn-C 12 H 26 decomposition in air at = 1.0 and several initial temperaturesT p;0 , and (d) net production rates of selected species as a function of reaction time for = 1.0 andT p;0 = 1200 K. . . . 37 4.6 (a) Ratio of the mass burning rate _ m u of a decomposed n-dodecane/air mixture at three representative initial temperature T p;0 values to that of un-decomposed mixture as a function of percentn-C 12 H 26 decomposed at = 0.7, 1.0, and 1.4. (b) Correlation between the mass burning rate enhancement and the ethylene yield in mass fraction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.7 Ranked logarithmic sensitivity coefficients of the mass burning rate _ m u with re- spect to (a) rate parameters and (b) binary diffusion coefficients for three reactant mixture conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.8 Ratio of u K ext of decomposedn-dodecane-air mixtures at several representative inital temperatureT p;0 values to that of un-decomposed reactant mixture (top pan- els) and ranked logarithmic sensitivity coefficients of u K ext with respect to binary diffusion coefficients for three reactant mixture conditions. . . . . . . . . . . . . . 42 5.1 Borghi diagram showing experimentally investigated fuels. Present work is shown indicated by red x’s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 5.2 Comparison of experimental and numerical temperature profiles at x = 1 mm above the jet exit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.3 Comparison of the velocity profiles measured experimentally and predicted nu- merically at x=1 mm above the jet exit. The vertical dashed line (right) indicates the edge of the pilot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.4 Axial (left) and radial (center and right) profiles of mean U and r.m.s. velocitiesu 0 . 52 5.5 Impact of Re jet on flame shape with constantT coflow =1500K andS L =11.5 cm/s . . . 53 5.6 Normalized flame intensity centerline profiles for various fuels at constantS L =11.5 cm/s at varying Re jet : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 5.7 Ignition Behavior of PPJB showing tip-quenched, tip-flickering, and tip-ignited states [40] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 5.8 Measured scaled flame height,H f /D jet , as a function ofS L forRe jet = 12,500 and Re jet = 25,000 (left and right panels, respectively). The profiles are for different fuels (see legend).) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 5.9 Adiabatic flame heights of Re jet =25,000 and 75,000 scaled by calculated T ad,jet andS L using the JetSurf mechanism . . . . . . . . . . . . . . . . . . . . . . . . . 57 viii 5.10 Impact of T coflow at a fixed Re jet (left) and Re jet at a fixed T coflow = T ad,jet (center) on the experimentally-measured flame heights. To emphasize the almost linear dependence on Re jet , a linear fit is shown in dashed lines. Comparison of H fl predicted by the LES and measured experimentally at the three different Re jet ’s for T coflow =1500K (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 6.1 Binarized flame brush vs. extracted flame perimeter used to calculatedS T,eff . . . . 64 6.2 Stitched PIV Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 6.3 Post-processing technique to determine instantaneous flame surface from Mie scat- tering images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 6.4 Nested PIV Example Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 6.5 Image showing burned vs unburned binary mask created from a particle density image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 6.6 Impact of Re jet on flame shape of Practical Fuels with ConstantT coflow =1500K and S L =15.3 cm/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 6.7 Centerline scaled intensity profiles of jet fuels . . . . . . . . . . . . . . . . . . . . 73 6.8 Flame heights (left) and maximum chemiluminescence intensities (right) of lean premixed jet flames adiabatic coflow and varying fuel type. The top two plots are atRe jet =25,000 and the bottom plots are atRe jet =50,000 . . . . . . . . . . . . . 73 6.9 Maximum CH* Mass Fraction for Various Fuels from 1D Laminar Flame Calcu- lations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 6.10 S T,eff for multiple fuels and correlation withH fl . . . . . . . . . . . . . . . . . . . 75 6.11 Binarized Average Flame Surfaces produced from Mie Scattering (dashed line) and CH* (white area) (left) and comparison of CH* intensity vs. c at the centerline (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 6.12 S T,GC; c=0:5 andS T,GC,CH* for Methane, Ethylene, and Jet A . . . . . . . . . . . . . 76 6.13 Axial Temperature Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 6.14 Radial Temperature Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 6.15 Stitched PIV Centerline Profiles. The pilot case represents ignited pilot and coflow but only air in the jet while the air case injects air in all three streams. . . . . . . . 80 6.16 Stitched 2D Turbulent Kinetic Energy Profiles . . . . . . . . . . . . . . . . . . . . 81 6.17 L int along the centerline and at a radial distance of x/D =15 . . . . . . . . . . . . 81 6.18 Ka andRe t calculated at the burner centerline . . . . . . . . . . . . . . . . . . . . 82 6.19 RescaledKa andRe t calculated at the burner centerline . . . . . . . . . . . . . . 83 6.20 Sample Projection of Local Velocity onto Flame Surface . . . . . . . . . . . . . . 84 6.21 S T,L as a function ofx=D (left) andU 0 as a function ofx=D . . . . . . . . . . . . . 84 ix 6.22 Abel-Inverted CH* Images of Jet Flames with inner and outer CH* cutoff for pro- jection in the black line and maximum intensity represented by the dashed line . . 85 6.23 S T,L as a function ofx=D (left) andU 0 as a function ofx=D for surfaces generated from Abel-inverted CH* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 6.24 Conditional Statistics ofU x andU 0 computed before the flame . . . . . . . . . . . 87 6.25 Conditional Statistics ofU x andU 0 computed after the flame . . . . . . . . . . . . 88 6.26 Probability Density Function (PDFs) ofU x andU 0 computed before and after the flame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 6.27 Ux 0 profiles for Sequentially Zoomed Windows . . . . . . . . . . . . . . . . . . . 90 6.28 S T,L as a function ofx=D (left) andU 0 as a function ofx=D . . . . . . . . . . . . . 90 6.29 Scaled Velocity, Temperature, and Chemiluminescence of Mean RANS Turbulent Jet Flame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 6.30 Flame Markers in RANS Turbulent Jet Flame . . . . . . . . . . . . . . . . . . . . 92 x Abstract An investigation of fuel and hydrodynamic effects is performed on piloted premixed jet flames. The investigation is carried out at varying laminar flame speed, varying heat losses, jet Reynolds number, fuel molecular weight, and fuel chemical classification. Well-characterized boundary conditions, well-resolved two-dimensional velocity fields from particle image velocimetry, and line-of-sight CH* profiles are analyzed. Global flame reactivity was analyzed from CH* and Mie Scattering images to determine global and local flame speed. Experimental results indicate that small amounts of heat losses may play a significant role on the jet reactivity as the flame heights scale with the heat loss from the jet. However, differences between flames with different fuels can still be seen in the absence of heat losses and these differ- ences are magnified at higher Reynolds numbers. Particularly, methane flames have consistently smaller global consumption speeds than ethylene for a given laminar flame speed and other fuels present approximately the same flame height at an intermediate value between the two extremes. A full characterization of velocity flowfield of the jet was performed using multiple Particle Image Velocimetry experiments. Highly resolved single camera PIV at the exit plane was able to characterize the boundary conditions at the exit of the jet, while stitched PIV allowed for the char- acterization of local turbulence properties interacting with the flame. The nested PIV configuration was used to resolve the entire energy spectra of vectors interacting with the flame by sequentially focusing in on smaller viewing windows until the Kolmogorov scale was resolved. Finally, edge detection techniques were developed for conditioned burned and unburned statistics. All of these measurements using fuels from methane to Jet A provide an experimental database of results that can be used as benchmarks for modeling studies of highly turbulent premixed flows. xi Chapter 1 Introduction 1.1 Background Since the industrial and technological revolutions, combustion has gradually become more and more essential for daily human life in applications such as transportation and electricity genera- tion. The International Energy Agency has assembled statistics which show that the total energy consumption in the world is rapidly increasing and in the past 40 years it has doubled from 4,667 Million Tons of Oil Equivalent (Mtoe) to 9,301 Mtoe. Oil consumption is still the majority, leading at 39.9%. It is followed by electricity consumption at 18% [1]. Figure 1.1 shows that even electricity generation relies mostly on fossil fuels. More than half of the energy used by the world is in the forms of oil and electricity which is sourced from carbon- based fossil fuels. Because the world is largely dependent on energy generation, it is important to study the source of the energy which powers cars, airplanes and power plants. The insight into the chemical kinetics, transport, thermodynamics and heat transfer can help us optimize the current modes of energy generation and push the limits of propulsion technology. 24 24 T R A N S F O R M A T I O N Electricity generation by fuel Hydro Other 2 Nuclear Fossil thermal 0 4 000 8 000 12 000 16 000 20 000 24 000 28 000 1971 1975 1980 1985 1990 1995 2000 2005 2010 2013 World electricity generation 1 from 1971 to 2013 by fuel (TWh) 1973 and 2013 fuel shares of electricity generation 1 6 131 TWh 23 322 TWh 1. Excludes electricity generation from pumped storage. 2. Includes geothermal, solar, wind, heat, etc. 3. In these graphs, peat and oil shale are aggregated with coal. Nuclear 3.3% Hydro 20.9% Other 2 0.6% Oil 24.8% Natural gas 12.1% Coal 3 38.3% Natural gas 21.7% Oil 4.4% Coal 3 41.3% Other 2 5.7% Hydro 16.3% Nuclear 10.6% 1973 2013 © OECD/IEA, 2015 Figure 1.1: Worldwide Electricity Generation [1] Due to the complex nature of combustion modeling, most designs continue to rely on empirical data. Combustion modes are application-dependent, so this study will focus on the applications to 1 the combustion phenomena that occurs in an aircraft engine. Most commercial and military aircrafts are powered by turbojet engines. A simplified descrip- tion of combustion in a turbojet engine begins with the air being sucked in through the air inlet and subsequently compressed through the compressor to create a pressure differential required for high thrust. To energize the flow, air is combined with fuel inside the combustor and burned. The exhaust exits the combustor in a continuous flow with no pressure buildup, as seen in Figure 1.2. The additional energy in the flow powers the turbine which is driven by the high temperature gases exiting from the combustor. Some of the energy generated in the turbine is fed back to power the compressor and other components necessary to maintain cooling. The heated exhaust gas is then accelerated through a converging duct then allowed to expand through an exit nozzle to produce thrust. Working cycle and airflow 15 Fig. 2-5-1 Airflow systems. Figure 1.2: Typical Turbojet Engine Structure and corresponding Pressure and Temperature Traces [2] Inside the combustor is where the energy is generated. The fuel is introduced into the system via an injector. In order to mix the fuel and air efficiently before burning, turbulence is generated in the dome/swirler which injects air perpendicular to the main flow to create a swirling flow field. A combustor needs to be tested and adjusted using empirical models across a wide range of conditions. A cutaway of a can-shaped combustor design is shown in Fig. 1.3. It shows where the flame is stabilized and the airflow path through the combustor. Jet A, JP-8 and JP-5 are all common examples of the hydrocarbon-based jet fuels used. The fuels are manufactured by distillation processes of fossil fuels and as a result, their composition may vary. They are blends of varying types of hydrocarbons with varying aromatic, olefin and paraffin content. The fuel is either injected as an atomized spray which is pre-vaporized, or premixed and pre-vaporized resulting in varying 2 types of flames. The igniter is placed in the combustion zone after the fuel and air have been rapidly mixed to create a swirl-stabilized, bluff-body stabilized, or piloted flame. This shear-layer stabilized device allows the combustor to operate at a wide range of conditions. For instance, it is able to consume leaner mixture ratios with lower temperatures without flame blowoff. Blowoff occurs when the flame is extinguished due to high velocity gradients in the flow which transport the radicals away from the flame zone. injected fuel droplets by rapidly bringing them to ignition temperature. 8. It is arranged that the conical fuel spray from the nozzle intersects the recirculation vortex at its centre. This action, together with the general turbulence in the primary zone, greatly assists in breaking up the fuel and mixing it with the incoming air. 9. The temperature of the gases released by combustion is about 1,800 to 2,000 deg. C., which is far too hot for entry to the nozzle guide vanes of the turbine. The air not used for combustion, which amounts to about 60 per cent of the total airflow, is therefore introduced progressively into the flame tube. Approximately a third of this is used to lower the gas temperature in the dilution zone before it enters the turbine and the remainder is used for cooling the walls of the flame tube. This is achieved by a film of cooling air flowing along the inside surface of the flame tube wall, insulating it from the hot combustion gases (fig. 4-4). A recent development allows cooling air to enter a network of passages within the flame tube wall before exiting to form an insulating film of air, this can reduce the required wall cooling airflow by up to 50 per cent. Combustion should be completed before the dilution air enters the flame tube, otherwise the incoming air will cool the flame and incomplete combustion will result. 10. An electric spark from an igniter plug (Part 11) initiates combustion and the flame is then self- sustained. Combustion chambers 37 Fig. 4-2 Apportioning the airflow. Fig. 4-3 Flame stabilizing and general airflow pattern. injected fuel droplets by rapidly bringing them to ignition temperature. 8. It is arranged that the conical fuel spray from the nozzle intersects the recirculation vortex at its centre. This action, together with the general turbulence in the primary zone, greatly assists in breaking up the fuel and mixing it with the incoming air. 9. The temperature of the gases released by combustion is about 1,800 to 2,000 deg. C., which is far too hot for entry to the nozzle guide vanes of the turbine. The air not used for combustion, which amounts to about 60 per cent of the total airflow, is therefore introduced progressively into the flame tube. Approximately a third of this is used to lower the gas temperature in the dilution zone before it enters the turbine and the remainder is used for cooling the walls of the flame tube. This is achieved by a film of cooling air flowing along the inside surface of the flame tube wall, insulating it from the hot combustion gases (fig. 4-4). A recent development allows cooling air to enter a network of passages within the flame tube wall before exiting to form an insulating film of air, this can reduce the required wall cooling airflow by up to 50 per cent. Combustion should be completed before the dilution air enters the flame tube, otherwise the incoming air will cool the flame and incomplete combustion will result. 10. An electric spark from an igniter plug (Part 11) initiates combustion and the flame is then self- sustained. Combustion chambers 37 Fig. 4-2 Apportioning the airflow. Fig. 4-3 Flame stabilizing and general airflow pattern. Figure 1.3: Combustor Cutaway and Flow Path [2] Due to the dynamic nature of the combustor physics and the different types of combustion taking place along the flow path, its challenging to model the behavior of a combustor. This study will assess the combustion process from first principles in order to try to understand how the fields of thermodynamics, chemical kinetics, fluid mechanics, and transport phenomena [3] collide in the field of turbulent combustion. The focus of this study is to examine how fuel effects interact with the high levels of turbulence in lean premixed turbojet engines so the focus will be on creating benchmark observables for experiments of combustion in shear layers. 1.2 Combustion Basics All combustion can be categorized as premixed or non-premixed combustion. Non-premixed 3 combustion occurs where the fuel and oxidizer mix and burn simultaneously in the reaction zone of the flame. Non-premixed flames occur in spray combustion, droplet combustion, candle flames and internal combustion engines. Therefore, the rate of combustion is limited by the rate of mixing. Because the reactants and products diffuse toward each other, they generally react at stoichiometric concentrations and the flame temperature is close to the maximum. Non-premixed combustion tends to be easier to stabilize. However, it is associated with higher pollutant formation, such as soot and nitrogen oxides that form in high temperature rich conditions. Also, non-premixed flames are prone to fuel leakage which may cause unburned hydrocarbons to be in the emissions. Premixed combustion occurs when the fuel and oxidizer are perfectly mixed before being ig- nited and then burned. The premixed combustion burning rate is determined by the flame speed, which is the speed of the propagating mixture into the unburnt reactants and a measure of the premixtures exothermicity, reactivity and diffusivity. Premixed flames can operate at fuel-lean or fuel-rich conditions with < 1 for the former and > 1 for the latter. is the equivalence ratio of the flame and is defined as: = F=O mix F=O stoich (1.2.1) where F=O mix represents the mixture fuel to oxidizer ratio in moles and F=O stoich represents the mixture ratio of a flame which is reacting in chemically balanced stoichiometric proportions. Lean premixed flames are considered more desirable due to reduced emissions since soot is virtu- ally non-existent in lean flames and the lower flame temperature produces fewer pollutants. The drawbacks of premixed flames may include flashback, flame blowoff and intermittent combustion. Flashback is the condition where the flame propagates backwards towards the igniter if the local velocity is too low which can cause damage to the combustor. Studies focusing on fuel-dependent properties of shear-stabilized lean turbulent premixed flames can offer insight into the stabilization mechanism of these flames and how to avoid blowoff, flashback and other potentially catastrophic phenomena. 1.3 Laminar Premixed Flames Locally, a turbulent premixed flame may behave similarly to a laminar flame and have a laminar flame structure. The simplest configuration that combines thermodynamics, chemical kinetics, fluid mechanics and transport phenomena is a laminar premixed flame following the ideal flame model [3]. The laminar flame speed is defined as the speed of the propagating flame front of a one-dimensional, adiabatic, steady, planar and unstretched flame [3]. The laminar flame speed is a fundamental property of the premixed flame and is sensitive to the chemical kinetics and molecular transport into the flame. It is a good measure of the flame’s exothermicity, reactivity, and transport properties. Therefore, it is a good metric to determine 4 how preheat temperature, pressure, fuel type, and equivalence ratio affect the fundamental flame structure. There are two main regions of the flame: the preheat zone and the reaction zone. The preheat zone is a balance between convection and diffusion and where the unburnt reactants are preheated up to a temperature at which reaction rates become significant. The reaction zone which is a where the reaction occurs and it is usually about a factor of 10 smaller than the preheat zone. The laminar burning velocity is given by the equation: S L /p ( n 2 1) q (=C p )e ( Ea 2RT ad ) (1.3.1) where S L (sometimes referred to as S 0 u ) is the laminar burning velocity, p is the pressure, n is the reaction order, is the mixture thermal conductivity,C p is the specific heat,E a is the overall activation energy, R is the universal gas constant andT ad is the adiabatic flame temperature. =C p represents the overall mixture diffusivity andE a =RT ad represents the reactivity of the mixture.S L is related to the mass burning rate f u by the equation S L = _ m u = u where u is the density of the unburned reactants. This is important because _ m u is the metric of most interest to propulsion devices since the overall thrust scales withf u . However, experimentally, _ m u is difficult to measure for certain experimental geometries and therefore S L is the parameter most often computed and compared. Another parameter that is often used to characterize laminar premixed flames is the flame thickness F . This value characterizes the balance of heat and mass into the flame front. F is defined as: F = (T max T u )=jrTj max (1.3.2) A laminar flame timescale can be defined from these two values as: t F = F =S L . t F charac- terizes the complex balance between convection and diffusion of heat and mass in the flame front. These three values are defined as fundamental properties of the premixed flame. 1.3.1 Flame Stretch S L is possible to achieve in theory, but in practice this type of flame is impossible to find. All real flames are affected by flame stretch, heat losses and confinement. Flame stretch,, is defined as the rate of area generation along the flame surface. = 1 A dA dT =r t v s;t + (V F n) (rn) (1.3.3) In this equation,r t represents the tangential gradient operator at the surface,v s;t is the tangential component of flow velocity at the flame,V F is the velocity of the flame, andn is the unit normal vector of the flame surface. A close examination of the equation reveals that the first term is 5 affected by flow nonuniformity throughv s . The first half of the second term is representative of experienced by a moving flame, and the second half represents the flame curvature since the gradient of the normal component will vanish without flame curvature. Therefore, is affected by aerodynamic straining (r t v s;t ), flame motion (V f n), and flame curvature (rn). Depending on the flame configuration, there are different predominant influences on the flame speed and burning rate. The differential diffusion of heat and mass into the reaction layer of the flame can interact with the local to produce stronger or weaker flames and its influence is strongly dependent on the flame geometry. The Karlovitz number is a non-dimensionalized stretch that measures the predominant flow effects on the flame given by the equation: Ka =t F =( F S L ) =( D r S 2 L ) (1.3.4) WhenKa approaches 0, the flame becomes the unstretched laminar case and whenKa is high it approaches limit phenomena such as extinction. This area generation from can also be caused by turbulence and the modification in area changes the burning rate and propagation rate of the surface. Turbulence can also modify this instantaneous area and increase the overall burning and propagation rate. The quantification of the area generation and modification in burning rate from turbulence has been long studied in the field of premixed turbulent combustion. 1.3.2 Fuel Structure and Breakdown in Laminar Flames As seen in the previous equation,S L is highly dependent on reaction rate of the flame which is dependent on fuel type and flame temperature. This is because different molecular structures have different reaction pathways and bond energies. Saturated hydrocarbons exhibit the lowestS L andT ad as well as the lowest average bond energy and tend to level off in maximumS L as carbon number increases. On the contrary, the maximum S L decreases with carbon number for alkenes and alkynes as the effect of the lone double or triple bond gets diluted by the rest of the saturated bonds within the molecule. This causes a decrease in the maximum S L toward the S L of n-alkanes. n-Dodecane, which is often used as a single component surrogate for jet fuel since jet fuel is mostly composed of n-parrafins has a S L very close ton-heptane so it can be assumed that all higher carbon numbern-alkanes converge on the sameS L . The fuel molecule is initially attacked through H-abstraction reactions by the H, O, and OH radicals. The large radical is then broken down into smaller hydrocarbon fragments according to the-Scission rule. The-Scission rule states that for a radical species, the unpaired electron at the radical site strengthens the adjacent bonds. Therefore, the C-C bond next to the radical-paired bonds are the most susceptible to break upon collision with another molecule. This causes even the heaviest ofn-alkanes to break up into small fragments mainly consisting of the H radical, methyl 6 radical, H 2 , CH 4 , and small alkenes mainly C 2 H 4 , C 3 H 6 , up to C 6 H 12 . A lumped reaction pathway and breakdown is given in Table 1.1. 438 Y. Gao et al. / Combustion and Flame 163 (2016) 437–446 Table 1 The skeletal fuel pyrolysis sub-model for n -dodecane. No. Reaction A n E a 1 n -C 12 H 26 = > C 6 H 12 + C 2 H 4 + n -C 3 H 7 + CH 3 8 .53 × 10 23 −2 .03 90,034 2 n -C 12 H 26 = > C 6 H 12 + 2 n -C 3 H 7 5 .64 × 10 26 −2 .68 88,171 3 n -C 12 H 26 = > C 4 H 8 -1 + C 2 H 4 + 2 n -C 3 H 7 7 .88 × 10 25 −2 .65 88,391 4 n -C 12 H 26 = > 2C 3 H 6 + 2 n -C 3 H 7 4 .00 × 10 26 −2 .66 88,392 5 n -C 12 H 26 + H = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 1 .30 × 10 6 2 .54 6756 6 n -C 12 H 26 + H = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 5 .20 × 10 6 2 .40 4471 7 n -C 12 H 26 + O = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + OH 2 .50 × 10 6 2 .40 5504 8 n -C 12 H 26 + O = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + OH 4 .60 × 10 5 2 .60 1768 9 n -C 12 H 26 + OH = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 O 1 .40 × 10 7 1 .80 974 10 n -C 12 H 26 + OH = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 O 4 .00 × 10 6 2 .00 −596 11 n -C 12 H 26 + CH 3 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + CH 4 9 .03 × 10 −1 3 .65 7153 12 n -C 12 H 26 + CH 3 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + CH 4 6 .00 3 .46 5480 13 n -C 12 H 26 + O 2 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + HO 2 4 .00 × 10 13 0 .00 50,930 14 n -C 12 H 26 + O 2 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + HO 2 1 .60 × 10 14 0 .00 47,590 15 n -C 12 H 26 + HO 2 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 O 2 4 .76 × 10 4 2 .55 16,490 16 n -C 12 H 26 + HO 2 = > C 6 H 12 + C 3 H 6 + n -C 3 H 7 + H 2 O 2 3 .80 × 10 4 2 .60 13,910 Note : Rate coefficient expressed as k = A T n exp (−E a /RT ). Units are mole, cm, s, cal, and K. developed from a version of JetSurF with lumped fuel cracking reac- tion steps [8] using directed relation graph (DRG) [9–11] , DRG aided sensitivity analysis (DRGASA) [12,13] , and linearized quasi-steady state approximations (LQSSA) [14] , and was applied in direct numer- ical simulations (DNS) of a turbulent counterflow n -dodecane spray flame [15] . Nevertheless, in the work of You et al. [4] , most of the combustion properties considered are those with negligible back mixing of burned or unburned mixtures. What was also not adequately studied was the near-limit phenomena, notably the extinction and ignition states in perfectly stirred reactors (PSR), which are typically turning points of the S -curve response [16] . In the present study, we inves- tigate the potential of using combined hybrid model development and model reduction to obtain highly efficient reduced models with less than 30 species for large hydrocarbon fuel combustion at high-T conditions. First, this hybrid model is assessed using the 24-species reduced model for n -dodecane and a variety of reactors and flames, including auto-ignition, PSR, premixed flame propagation, and extinction of premixed and non-premixed counterflow flames. The validity of the assumption of fast fuel cracking at high temperature conditions is subsequently investigated with 2-D DNS of a lean turbulent premixed flame for n -butane/air using a reduced model derived from the detailed USC-Mech II. 2. Performance of a reduced model for n -dodecane with lumped fuel cracking steps 2.1. A brief review of the reduced model An approach to lump the fuel cracking reactions has been demon- strated in [4,8] . This approach assumes that the intermediates of the fuel cracking are all in quasi-steady state. It reduces the chemical complexity of the reaction processes and the model size in an effi- cient manner. For n -dodecane, the resulting lumped model contains only three species, namely, n -dodecane, 1-pentene and 1-hexene and approximately one and a half dozen reactions. The rate coefficients of the lumped reaction model are based on JetSurF 1.0 [17] . The lumped pyrolysis model was combined with USC-Mech II [7] to obtain a com- plete model of 123 species and 977 reactions, referred to as JetSurF 1.0-l, which has been validated against the detailed model and ex- periments over a wide range of conditions [8] . Subsequently, this lumped-detailed reaction model was reduced to 24 species and em- ployed in a study of spray combustion of n -dodecane in a turbulent counterflow flame [15] . The three steps taken in the above model reduction are described here. The DRG-based methods, including DRG and DRGASA, were employed to remove unimportant species and reactions first from a range of combustion responses over the pressure range of 1 to 10 atm, initial temperatures from 10 0 0 to 160 0 K for auto-ignition, inlet tem- perature of 300 K for PSRs, and equivalence ratios from 0.5 to 1.5. The H atom was selected as the starting species in the graph search- ing in DRG and the obtained skeletal model consists of 47 species and 359 reactions. After skeletal reduction with DRG, the model was further reduced with DRGASA to obtain a smaller model. The worst case error in target parameters induced by removing a single species was first tested and then sorted in ascending order. The species were then eliminated one by one until the worst case error in the tar- get parameters reaches the given error tolerance. In DRGASA, auto- ignition delay with initial temperature from 10 0 0 to 1600 K and ex- tinction residence time of PSRs with inlet temperature of 300 K over the pressure range of 1 to 10 atm and equivalence ratio from 0.5 to 1.5, were selected as the target responses with an error tolerance of 20%. The final skeletal model is comprised of 193 reactions and 31 species: N 2 , H, O, OH, HO 2 , H 2 , H 2 O, H 2 O 2 , O 2 , CH 2 , CH 2 ∗ , CH 3 , CH 4 , HCO, CH 2 O, CH 3 O, CO, CO 2 , C 2 H 2 , C 2 H 3 , C 2 H 4 , C 2 H 5 , C 2 H 6 , CH 2 CHO, aC 3 H 5 (allyl), C 3 H 6 , n C 3 H 7 , C 2 H 3 CHO, C 4 H 8 -1 (1-butene), n C 12 H 26 ( n - dodecane), and C 6 H 12 (1-hexene). Reactions of any eliminated species were also removed. The skeletal sub-model for fuel pyrolysis is pro- vided in Table 1 . In the last step, this skeletal model was further re- duced with the LQSSA method [14] . Seven global QSS species, namely CH 2 , CH 2 ∗ , HCO, CH 3 O, C 2 H 3 , C 2 H 5 , and n C 3 H 7 , were identified by excluding all species with nontrivial projection to the slow chemical subspace, thus resulting in a 24-species model. The QSS species are eliminated from the transport equations and can be solved with a set of internal algebraic equations. The QSSA equations are evaluated an- alytically using a graph-based method [14] to ensure high accuracy and robustness. 2.2. Extended model validation Validation of the reduced model is rather limited in Ref. [15] and is extended in the present study, particularly regarding the effects of fuel cracking. The reduced model is first compared with the detailed JetSurF 1.0 and the lumped-detailed model, in which the elementary fuel cracking reactions of JetSurF 1.0 are replaced by lumped semi- global steps, for ignition delay and PSR extinction over a wide range of pressures, temperatures and equivalence ratios. Figure 1 shows ex- cellent agreement of the three models despite some minor discrep- ancies between the detailed and lumped-detailed models. The normalized total mass fraction of the species with four or more carbon atoms, denoted by C 4 + , are plotted in Fig. 2 as a function of temperature in auto-ignition and PSRs for different equivalence Table 1.1: Lumped Mechanism for High-Temperaturen-Dodecane Oxidation [4] Since the same composition of hydrocarbon fragments are ultimately entering the flame front, the flame speeds are very similar for heavy hydrocarbons. A new lumped model for jet fuel offers insight into why this might be [4]. This mechanism which has been validated against extinction strain rates, flame speeds, and reactors shows very good agreement for all these values using a 24 species model. This paper uses USC Mech II and H 2 and C 1 -C 4 kinetics as the backbone and assumes the high temperature oxidation and subsequent breakdown into C 1 -C 4 happens very quickly and these molecules are what ultimately are consumed in the reaction zone. This may not be the case for high levels of mixing and high strain and stretch rates which are seen in turbulent flame mixing and reaction layers which will be discussed later. 1.3.3 Molecular Transport and Heavy Hydrocarbons For premixed flames, the Lewis numberLe is an important parameter that responds strongly to stretch effects and it is defined as: Le = D d (1.3.5) 7 where is the thermal diffusivity of the reactant mixture and D d is the mass diffusivity of the deficient reactant which is the fuel for lean mixtures and oxidizer for rich mixtures. Local or global differences in diffusive vs thermal transport can cause the stretched flame speed to differ significantly from the unstretched flame speed. This is due to the balance of heat transfer out of the flame and reactants toward the flame which is described byLe. Since diffusivity is inversely proportional to the square root of molecular weight, the mixture averagedLe can be approximated as: For< 1;Le = N 2 D Fuel = s MW Fuel MW N 2 (1.3.6) For> 1;Le = N 2 D O 2 = s MW O 2 MW N 2 (1.3.7) When the mixture is rich, theLe remains fairly constant and slightly greater than 1 for fuel/air mixtures. Le> 1 for a lean heavy hydrocarbon/air mixture andLe< 1 for a H 2 /air mixture. An- other factor that occasionally comes into play is the differential diffusion effect which is prevalent in heavy hydrocarbons. The differential diffusion represents the difference between the mass dif- fusivity of fuel and oxidizer into the flame. The mass diffusivity decreases with increasing carbon number and therefore the ratio of D O 2 =D Fuel becomes increasingly important for heavy hydro- carbons and mass transport of the oxidizer may preferentially be transported into the flame which may cause the flame to locally become more lean/rich and modify the local reactivity. For fuel rich heavy hydrocarbon mixtures, this will cause the flame to get closer to stoichiometric and enhance the local flame temperature. All practical flames are stretched and the counterflow flame (which will be discussed next) experiences positive stretch while the bunsen flame which is commonly used for laminar flame speed measurements experiences negative stretch. 1.3.4 Counterflow Flames One of the most important flame configurations that combines all of the concepts mentioned in the previous sections is the laminar counterflow flame. The laminar counterflow flame config- uration is one of the standard methods used for measuring stretched laminar burning velocity and then extrapolating to zero stretch to approximate the laminar flame speed. It is created when two round jets which are flowing equal momentum jets impinge onto a stagnation plane. Premixed counterflow flames either consist of a premixed fuel/air jet impinging on an inert N 2 gas plane, a vitiated air mixture or a twin adiabatic flame. The flow is reduced to a quasi-1D flow when a streamfunction is defined to reduce the solution 8 such that the flow velocity is a function of x only using the equations: (x;r) =r 2 U(x);u = 1 r @ @r ;v = 1 r @ @x (1.3.8) This is inserted into the continuity equation: = @ @x (ru) + @ @r (rv) = 0 (1.3.9) This is inserted into the momentum equation. Additionally, given the low-Ma assumption, the radial pressure gradient is also a constant eigenvalue of the flow with the equation: 1 r @P @r =C (1.3.10) The counterflow configuration has a continuously varying stretch and is expressed at the centerline as: = du dx (1.3.11) This approximation makes this configuration convenient to model. The flow velocity along the centerline first decreases as the flow approaches the stagnation plane and then rapidly accelerates when it hits the flame due to thermal diffusion and thermal expansion. The minimum velocity right before the flow acceleration is called the reference flame speedS u;ref . This value is chosen for experimental convenience. However, the actual stretched flame speed is the velocity just before the temperature rise begins to occur whileS u;ref occurs in the preheat zone. The reference stretch is chosen in a similar manner as the maximum value in the velocity gradient before the flame zone since the value in the unburned region is easier to measure than the burned region. Extinction of the counterflow flame is a very userful metric for determining reactivity of the flame and is dependent on both stretch and Lewis number so it is a parameter highly sensitive to fuel type, reactivity, and velocity. Extinction can occur for two separate reasons dependent onLe of the reacting mixture. Due to the transport asymmetry in the counterflow configuration in which the diffusive transport occurs normal to the flame surface while the convective transport can occur radially following the direction of the streamlines. This causes mass to be diffused from external streamlines to the flame and heat to be diffused from the flame to external flow. ForLe > 1, like for lean premixed flames with heavy hydrocarbons, heat loss from the flame dominates over the enthalpy gained from mass transport into the flame and therefore, the burning rate will be reduced with increasing stretch. WhenLe< 1, for example hydrogen and methane/air flames, the burning rate increases with increasing stretch and therefore, the burning rate will continue to increase until the flame gets pushed closer and closer to the stagnation plane. The flame thickness will begin to decrease and the residence time causing incomplete reaction and eventually leading to extinction. 9 1.4 Turbulent Premixed Combustion 1.4.1 Basics of Turbulence Since a premixed flame will be influenced by the turbulence of the unburned reactant mixture into which it is propagating, the non-reacting turbulence properties will first be defined. Therefore, attempts will be made to first understand and characterize the unburned nature of the flowfield although there is feedback between the flame and turbulence due to the large gradients of density, temperature and species in the reaction layer. Since turbulence is by definition random in nature, attempts to characterize turbulence are not exact but statistical in nature. If a point in the mixture is ensemble-averaged there are two main properties of the flow: the mean velocity u and the fluctuating component of velocityu 0 : u 0 = ( <u i u i > 3 ) 1=2 = (2k=3) 1=2 ) (1.4.1) whereu i is the i component of the 3D velocity vector andk is the averaged turbulent kinetic energy of the flow. The dissipation rate from this value is then defined as: " =h @u i @x j @u i @x j i (1.4.2) The characteristic length of the flow is usually proportional to the largest scale of the flow. In the case of a jet, it is the jet diameter. This length scale at which energy is introduced into the flow is called the integral length scale, L 0 and it is defined statistically based on measured flow properties. It is defined as: L 0 =u 03 =" (1.4.3) According to Kolmogorov’s 1941 theory on turbulence [5], energy is transferred from the largest to smallest scales according to the energy cascade in which turbulent eddies break down into smaller and smaller vortices until being dissipated at the order of the viscosity of the flow. The lengthscale where the turbulent dissipation occurs is known as the Kolmogorov scale given by: = ( 3 " ) 1=4 (1.4.4) Further time and velocity at the Kolmogorov scale are given by: t = ( " ) 1=2 ;u = (") 1=4 (1.4.5) The energy transfer per eddy turnover time of the largest scales has to equal the dissipation at the 10 smallest lengthscales. Therefore, the following definitions will be used for dissipation: " = u 03 L 0 = u 3 (1.4.6) The Taylor length scale is defined as an intermediate scale between the Kolmogorov scale and integral scales and replaces the averaged gradient value in the definition of the dissipation byu 0 =. This leads to the definition: " = 15 u 02 2 (1.4.7) The factor of 15 comes from the considerations seen in homogeneous isotropic turbulence. is then defined from this equation as: = (15u 02 =") 1=2 =u 0 t (1.4.8) This equation reveals that a physical interpretation of is the distance that a large eddy convects a Kolmogorov eddy during its turnover timet . The measure of the r.m.s strain rate based on these definition then is: u 0 = =t 1 =(15) 1=2 (1.4.9) 1.4.2 Turbulence in a Free Shear Flow The experimental configuration chosen was an axisymmetric turbulent round jet which has been characterized well for the non-reacting case. The advantage of the configuration is that it is thoeretically statistically dependent only on the axial and radial coordinates and not dependent on angle and therefore is an attractive configuration to study for 2D velocity profiles [6]. Since it is a statistically stationary flow, it is subject to statistical analysis of its major moments. The first moment is the mean velocity. For analysis along axial distance, the centerline velocity is defined asU 0 (x) =U(x; 0; 0) and jet half width defined as the radial distance from the centerliner 1=2 (x) in which < U 0 (x;r 1=2 (x); 0) >= 0:5U 0 (x). The jet continually spreads with increasing axial distance and exhibits self-similarity since the shape of the profile does not change when plotting radial profiles of scaled U=U 0 versus r=r 1=2 , causing all velocity profiles to collapse on a single curve. This holds only for the region outside of the developing region of the jet which can be up to x=D = 2530 depending on the flow but the velocity profiles can still be scaled by this parameter to see how well the self-similarity holds for different fuels. For a non-reacting round jet, it can be shown from experiments thatU jet =U 0 (x) increases lin- early withx=D, the slope of which is a constantB and dependent on a virtual origin atx 0 . This corresponds with a definition of a linearly increasingr 1=2 (x) the slope of which is defined as the 11 spreading rateS according to the following equation:r 1=2 (x) =S(xx 0 ) [6]. The definition of a linearly decreasing U 0 and a linearly increasing r 1=2 (x) effectively can- cel each other out when calculating the local Reynolds number along the centerline Re 0 (x) = r 1=2 (x)U 0 (x)= so that it is also independent of x. B and S are shown to be independent of Reynolds number for various measurements and are approximatelyS = 0:094 andB = 5:8 ac- cording to various experiments. The Reynolds stresses and r.m.s velocity are also self-similar. Panchapakesan and Lumpley [7] found that after the development regionu 0 0 =U 0 (x) decays linearly as a function of x with a slope of 0.25. A turbulent viscosity T can be defined as: <uv>= T @ <U > @r (1.4.10) These self-similarities in behavior are important to keep in mind when analyzing the behavior of a reacting turbulent free jet to understand how it changes the typical downstream jet behavior. 1.4.3 Regimes of Premixed Turbulent Combustion The different regimes of turbulent combustion can now be defined based on how different lengthscales of turbulence will interact with the premixed flamefront. The turbulent Reynolds number can be defined from the previous quantities as: Re t = u 0 L 0 (1.4.11) With a couple assumptions, we can correlate the flame lengthscales with the turbulent Reynolds number [8]. Assuming the Schmidt number =D r is close to unity, can be replaced with D r where D r is the diffusivity of the reactant. Assuming Le 1 as well, we can assume that the flame thickness F is proportional toS L and the diffusivity of the reactant mixtureD r giving the relationship F =D r =S L . Using this information, we can redefineRe t as: Re t = u 0 L 0 S L F (1.4.12) An additional parameter is necessary to relate the turbulence to the flame speed and flame time. The turbulent Damk¨ ohler numberDa of the timescale of the largest eddies of the flow is defined as: Da =t =t F =S L L 0 =u 0 F (1.4.13) As mentioned earlier,Ka is a measure of the non-dimensional imposed strain rate onto the flame surface. The turbulent Ka on the order of the Kolmogorov scales of the flow can therefore be 12 defined as: Ka =Da 1 =t F =t F =t = 2 F = 2 =u 2 =S 2 L (1.4.14) Ka is the inverse ofDa at the Kolmogorov scale. A common theme emerges: there are two main scales of note that influence the burning rate of the flame. One is the relationship between the convective flow scales to the propagation speed of the flame represented by the relationshipu 0 =S L . Another important scaling parameter is the ratio of the flame thickness to the size of the eddies in the flow which is represented by the relationship L 0 = F . These parameters are related using the previous equations forKa andRe t : u 0 =S L =Re t (L 0 = F ) 1 =Ka 2=3 (L 0 = F ) 1=3 (1.4.15) With the use of this equation, a regime diagram can be created relatingu 0 =S L andL 0 = F where the linesRe t = 1,Ka = 1 andKa = 1 (Ka of the reaction zone thickness which is approximately 1/10 of the F ) represent divisions between different regimes of premixed turbulent combustion. This diagram was first proposed by Borghi and modified by Peters [9, 10] and is shown in Fig. 1.4. 0.1 1 10 100 1000 0.1 1 10 100 1000 Laminar ,lames Re t =1 u'=S L Wrinkled ,lamelet Corrugated ,lamelet Thin reaction zone Broken reaction zone Ka=1 Ka δ =1 u'/S L L 0 /δ F Figure 1.4: Regimes of Premixed Turbulent Combustion A couple observations are immediately apparent. First, theRe t = 1 line represents the division between laminar and turbulent flames which have Re t > 1. Now, the regimes of wrinkled and corrugated flamelets are characterized by Ka < 1 which represents fast chemistry compared to the flow timescale andRe t > 1 which places it in the regime of turbulent flow. The eddies in this regime merely wrinkle and stretch the flame which still reacts in thin laminar flamelets. The eddies will wrinkle and stretch the flame on a global basis but cannot penetrate the reaction layer of the 13 flame. The thin reaction zone regime is characterized by eddies that are small enough to penetrate the preheat zone of the flame but do not modify the reaction zone of the flame i.e. < < F . Finally, the broken reaction zone occurs when the smallest eddies can modify the reaction layer of the flame by transporting local radicals into or away from the flame and locally extinguish the flame. It should be emphasized that this diagram is based on assumptions of Le=1 and is used as a basis to understand different flame types. The Gibson scale is defined as the lengthscale of the eddies that interact locally with the turbulent flamefront and is different dependent on the turbulent properties of the flow [11]. It is defined as: L G =S 3 L =" =L 0 (S L =u 0 ) 3 (1.4.16) There are two ways to decrease the Gibson scale of the flow to push the flame closer to the extinction limits. One is by increasing the u 0 and the other is by decreasing the reactivity of the flame S L . The regimes are merely a guideline and it does not follow that all flames with the same turbulent properties will have the exact behavior as mentioned by the regime diagram. Turbulence is statistical in nature and there is feedback between the locally stretched burning rate and turbulent properties. Based on this diagram, the interaction between turbulence and chemistry is most enhanced in the thin and broken reaction zone regimes since the breakdown of heavy hydrocarbons happens at the scale of the preheat and reaction zone thickness. It is precisely these regimes which are the focus of this study to try to isolate fuel effects on premixed turbulent flames. A couple observations can be made from the regimes of turbulent combustion and scaling laws. The first is that all of the turbulent combustion scaling laws are based on averaged unburned mixture turbulent properties. The only dependence on flame properties comes from the laminar flame and is based onS L and F since the local flame front is still dependent on these properties. Therefore,S L will be kept constant when comparing between different hydrocarbon fuels. 1.4.4 Turbulent Burning Velocity Many attempts have been made to correlate turbulent properties with the increase in mass burn- ing rate that is often seen in turbulent flames. The simplest and most instructive of the approaches is detailed in the innovative work by Damk ohler [12] in which he assumes an instantaneous wrinkled turbulent flame surface which he assumes locally propagates at the laminar burning rateS L . The increase in mass burning rate is attributed to the increase in wrinkled flame area produced by the turbulent fluctuations in the equation: _ m = u S L A T = u S T A (1.4.17) 14 whereS T is the turbulent flame speedA T is the area of the turbulent wrinkled surface andA is the cross sectional duct area. This gives rise to the relationship: S T S L = A T A (1.4.18) Assuming that the increase of the wrinkled flame surface area is proportional the the increase of local flow velocity overS L gets the following relationship: S T S L = A T A = S L +u 0 S L = 1 + u 0 S L (1.4.19) Whenu 0 >>S L thenS T u 0 When the turbulence is on a smaller scale, the turbulence will modify the transport in the reaction zone and the turbulent diffusivity will take precendence over the laminar diffusivity giving the relationship: S T S L r D T D L = r u 0 S L L 0 F (1.4.20) Both equations rely on theu 0 =S L scaling relationship to determineS T so different research groups have attempted to create an empirical correlation withS T =S L andu 0 =S L with the equation: S T =S L = 1 +C( u 0 S L ) n (1.4.21) Again, this correlation for S T factors into the justification for using a constant S L to compare between different fuel/air mixtures. A further of experimental techniques to measure turbulent burning velocity will be discussed in Chapter 3. 1.4.5 Fuel Effects on Turbulent Premixed Combustion Turbulent combustion has been largely studied over the years for hydrogen and methane [13, 14], and valuable insight has been gained. However, in applications such as high-speed air-breathing propulsion, large molecular weight liquid fuels (C 5 and above) are used in the presence of intense turbulence [15]. Heavy hydrocarbons have in general, a lower resistance to cracking and they typ- ically decompose endothermally into smaller fragments when exposed to the high temperatures within the preheat zone [15–18]. Additionally, the diffusivities of heavy fuels are distinctly dif- ferent compared to hydrogen and methane. It is of interest to note that fuel effects have attracted relatively less attention in past turbulent combustion studies (e.g., [19–23]) and that such effects are lumped in most cases under one fundamental flame property the laminar flame speed (S L ). Under high levels of turbulence and high Karlovitz number, Ka, a regime of turbulent com- bustion is reached where the smallest eddies of the flow can enter and broaden the preheat zone 15 and modify the flame structure [13, 19, 24–26], labeled as the thin reaction zone regime by Borghi [10] and Peters [9]. The broken reaction zone regime corresponds to flow conditions under which even smaller eddies are able to penetrate the reaction zone [9, 10]. Though established in theory, conflicting experimental studies have been presented on how the preheat zone is modified in the thin reaction zone regime. With increasing turbulence intensity for instance, Dinkelacker et al. [24] reported flame-thinning trend, while Chen and Bilger [26] reported a broadening of the flame thickness. In order to investigate finite-rate chemistry effects, Bilger and coworkers [27, 28] de- veloped a piloted premixed jet burner (PPJB) that allows for the stabilization of highly turbulent premixed jet flames in the thin and broken reaction regimes. There is experimental evidence that the differential diffusion effects, which dominate the inter- actions in laminar flames for high carbon number fuels close to extinction [15, 18], are important in turbulent combustion [29]. The effects on the turbulent flame speed in fan-stirred bombs in the flamelet regime of combustion, for example, show a dependence on differential diffusion and Lewis number similar to what can been seen in stretched laminar C 0 -C 3 hydrocarbon flames [30–32]. However, it is not clear if the same correlations apply to highKa flows where turbulent trans- port dominates. There, small eddies can broaden the preheat zone and thicken or even break the reaction zone [26, 33, 34]. The impact on the chemical processes is still not well characterized, especially for heavy hydrocarbon fuels. If intense turbulence modifies the preheat zone structure as shown in previous studies, then it is expected that an attendant effect on the fuel temperature-time history could in turn affect the supply of reactants into the reaction zone. More specifically, if the thermal structure of the preheat zone is modified by turbulence, then the effects on methane could be minor given its greater resistance to decomposition. However, this may not apply to heavy hydrocarbons that can decompose readily. For example, a prolonged exposure to high temperature within the preheat zone due to broadening, could modify notably the type of hydrocarbon fragments that are supplied to the reaction zone. Very few standard burner designs allow for high Reynolds number premixed flame stabiliza- tion. Among them, the Piloted Premixed Jet Burner (PPJB) approach was first introduced by Bilger and coworkers [27, 28] in order to investigate finite-rate effects for near-limit methane/air flames. The PPJB configuration is adopted in the present study that involves both experiments and computations in order to assess fuel effects in turbulent premixed jet flames. Experiments in the presence of high levels of turbulence are useful for understanding global and local quantifiable flame parameters. However, it may be useful to also consider a more canon- ical configuration to understand the effect of different flow length and time scales on properties of a propagating flame front. V ortices of a wide variety of length and time scales are gener- ated naturally in various turbulent flow fields. As such, turbulent combustion can be described as the modification, production, and dissipation of flame surface by cascading vortices of various 16 wavenumbers [35]. 1.4.6 Objectives and Structure of Thesis Based on the aforementioned considerations, the main goal of this investigation was to provide experimental evidence of potential fuel effects on turbulent combustion under high levels of shear and turbulence by considering light gaseous and heavy liquid hydrocarbons of various chemical classifications in an atmospheric pressure piloted premixed jet burner (PPJB) [27, 28] facility. More precisely, the objectives are two-fold. The first objective is to generate well-characterized experimental data for a variety of fuels from small alkanes/alkenes (methane/ethylene) to large n-alkanes (propane and n-heptane) to aromatic fuels (toluene). The second objective is to identify possible fuels effects under a fixed and identicalS L . These objectives will be achieved by using PPJB experiments combined with matching RANS Simulations in order to qualitatively compare flame behaviors. Any possible fuel effects will be isolated by conducting the experiments with fuel/air mixtures with the exact sameS L . Additionally, stretched laminar flame speeds will provide further insight on the fuel effects on on propagation speed of laminar flamelets with modified reactant compositions due to fuel pyrolysis and stretch. The global experimental observables are flame height and luminosity, as well as two-dimensional spatially and time-resolved velocity profiles that allow for the determination of key turbulence properties. A key measurement that is lacking in the field of turbulent combustion is turbulent combustion velocity statistics before and after the flame. Particle Image Velocimetry (PIV) mea- surements of the flame will be attempted in various forms in order to resolve the differing scales of the jet flame in time and in space. Additionally, some numerical studies on fuel decomposition in the presence of laminar stretch and vortex/flame interactions will be examined. The organization of this document is as follows: Chapter 2 is dedicated to the description of the experimental setup and Chapter 3 describes the numerical approach. Chapter 4 presents numerical investigations of stretched premixed counterflow flames of partially reacted mixtures of n-dodecane to understand the effect of decomposition on fundamental flame proprties. Chapter 5 presents global and local PPJB flame observables in ignited coflow with liquid and gaseous fuels of different chemical classification and Chapter 6 presents results performed on Jet A to understand how practical fuels perform in this basic setup. 17 Chapter 2 Experimental Approach 2.1 Burner Setup Turbulent combustion has been studied in a wide variety of geometries and attempts to correlate experimental values attained in this wide variety of geometries (spherical, Bunsen, etc) with the same turbulence properties have been largely unsuccessful. This is indicative of the fact that the turbulent wrinkling process is geometry-dependent and contains memory of the wrinkling in an upstream location of the flame [13]. For example, a turbulent bunsen flame which is stablized at the base has a smaller value ofS T at the rim than the tip region. Due to evidence that realistic flames undergo geometry-dependant wrinkling and therefore burning rate, it is helpful to categorize the different flame types. Driscoll describes the categories of flames as: 1. Envelope flames (including Bunsen, jet, and other axisymmetric configurations) 2. Oblique flames (V-shaped flames, bluff-body stabilized flames) 3. Flat flames (low-swirl, counterflow, or diffusion burners) 4. Spherical flames which include fan-stirred chambers This experiment is focusing on the first type of flame: the envelope flame configuration, more specifically a premixed jet flame. The main goal of this burner design was to achieve measure- ments of turbulent premixed flames at high Reynolds numbers and high turbulence intensities. High levels of turbulence intensity can be created in axisymmetric turbulent jet flames and it is a relatively canonical configuration for modeling. For this reason, an axisymmetric turbulent jet flame configuration was chosen. Envelope flames are still used widely to study highly turbulent combustion. Turbulent bunsen flames have also been used to measure instantaneous flame shapes and global turbulent burning velocities by visualizing the 2-D instantaneous flame contours. To stabilize even higher exit velocity flames and generate higher turbulence intensities requires an additional flame anchoring mechanism. A lower velocity pilot flame is often used for this purpose. Piloted bunsen flames have been studied by Chen and Bilger [36] and have revealed potential non-unityLe effects 18 on the local burning rates of the flame due to the mean flame curvature. However, this configuration is not able to reach the lowDa and highKa necessary to broaden the flame front sufficiently and reach the broken reaction zone regime since it is subject to heat loss and entrainment of ambient air away from the pilot flame. The advantages of the atmospheric piloted turbulent premixed jet flame configuration are as follows: Axisymmetric boundary conditions with easier potential for modeling Mass flow rate through the cone is known (for global consumption speed measurements) Goods optical acces to the flame for laser-based techniques since no chamber is required High levels of turbulence can be generated in the shear layer between the flows Flow is statistically stationary so it is easy to attain statistics of the average, root mean squared (r.m.s), and higher order moments of the flame Fuel flexibility of the central jet However, the disadvantage is that the turbulence is not homogenous and isotropic and although the burner is axisymmetric, it has three-dimensional effects which must be quantified and may be less than ideal if there are flow non-uniformities. Also, there are dilution effects which increase with flame axial distance as the flame will mix with the surrounding air or vitiated co flow. Never- theless, shear-generated premixed turbulent layers stabilize turbulent jet flames in turbojet engines and there is merit to studying the properties which can be measured and modeled in these shear layers. The present work relies on the original PPJB burner design [27, 28] with modifications to allow studies of fuels with a wide range of molecular weights. A schematic of the burner design used in experiments is shown in Fig. 2.1. The burner consists of two co-annular premixed flames surrounding and stabilizing the high exit velocity central jet. A three-stream flow with a large diameter coflow was created so that the central jet flame would be able to experience adiabatic flame temperatures at the flame tip and avoid the quenching that occurs as the influence of the pilot, which is strong close to the nozzle becomes diluted by the outside air at higher axial distances. The two-stream pilot and jet flow is useful for studying heat loss and quenching effects on the reactivity of jet flames. The main jet consists of a 900 mm long smooth straight stainless steel tube with an internal diameter ofD=5.84 mm and an external diameter of 6.35 mm. KeepingS L constant, the main jet exit velocityU jet and fuel are varied to investigate finite-rate chemistry and fuel effects. 19 Figure 2.1: Piloted premixed jet burner (PPJB) cutaway. All measurements are in mm. The main jet is surrounded by a pilot flow coming out of a co-annular tube with internal di- ameter ofD pilot =22.9 mm and outer diameter of 24.1 mm. The nozzle edge of the jet is located 0.6mm above the pilot nozzle. A small step (0.5 mm thick) located 11 mm from the burner exit is used to hold a 5 mm thick brass perforated plate. This plate is composed of 52 holes (1.52 mm in diameter) and serves as anchor for the stoichiometric premixed methane/air flames. The pilot tube has an unburnt exit velocity of 0.75 m/s before the step. The pilot tube is 70 mm in length and externally conical in shape with a 32.4 mm diameter base. Finally, the outermost co-axial coflow uses hot products to insulate thermally the jet. Re jet = U jet D u (2.1.1) whereU jet is the exit velocity of the jet which varies from 50-250 m/s and u is the viscosity of the reactant mixture. A new fuel vaporization system was designed based on existing technology developed for heavy hydrocarbon investigations in laminar flames [15, 18]. This system has a much higher flowrate capable of fully vaporizing up to 20 mL/min of liquid fuel with air flow rates of up to 400 SLPM. Fig. 2.2 shows that the system consists of inline heaters that preheat the air which issues into heated stainless steel chambers. The fuel is then injected in a crossflow configuration into the chambers with a fine mist of droplets from the tip of a Meinhardt liquid nebulizer [15]. The fuel is held in a Chemyx syringe and the flow rate controlled with a high accuracy Nexus syringe pump. Special care has been taken in order to avoid hot spots that can cause fuel decomposition and cold spots that can cause fuel condensation throughout the flow system [18]. The gaseous flow rates are controlled accurately using sonic nozzles. 20 Temperature Controller Thermocouple Temperature Controller Thermocouple Air Compressor Test Gauge Test Gauge Fuel Cylinder Pressure Regulator Pressure Regulator Sonic Nozzles Sonic Nozzles In-Line Heater Temperature Controller Thermocouple To burner Water inlet Water outlet Temperature Monitor Thermocouple Syringe Pump Nebulizer Test Gauge Pressure Regulator Heating tape Heating tape Figure 2.2: In-House Vaporization Diagram 2.2 CH* Chemiluminescence One of the easiest metrics to observe with a simple camera setup and is even visible to the naked eye is the global flame shape. The chemiluminescence of hydrocarbon flames is due to OH* emitting around 307 nm, CH* emitting around 431 nm, C 2 * swan-bands between 430 nm and 650 nm, and background CO 2 * with a temperature-dependent broadband emission spectrum from 300-600 nm [37–39]. In laminar flames, the flame chemiluminescence signal is a relatively good marker of the position of the maximum heat release rate (HRR max ) [39]. The spatial overlapping is confirmed by comparing the normalized mole fraction profiles of CH*, OH*, and CO 2 * to the more traditional flame markers such as OH, CH 2 O, and their product, as well as temperature and heat release rate obtained through modeling of freely propagating flames which is discussed in Carbone et. al [40]. Additionally, under laminar conditions, the chemiluminescence signal can be a good marker of HRR max [37, 39] as depicted in Fig. 2.3 in which the peak mole fractions of the chemiluminescent species (CH*, OH*, and CO 2 *) are shown as functions of HRRmax and are compared to the peak of the product of OH and CH 2 O mole fractions. The results suggest that the correlations between the peak mole fractions of the chemilumi- 21 Figure 2.3: Computed correlations of key flame markers against the heat release rate obtained by laminar flame modeling. nescent species andHRR max do depend on the fuel type and exhibit a clear positive slope under most conditions. Similar trends can be seen for the correlations for the peak of the product of OH and CH 2 O mole fractions. Those results can cause ambiguity regarding the interpretation of chemiluminescence measurements since its presence clearly indicates vigorous burning whereas its absence does not necessarily suggest local extinction especially in turbulent environments that involve complicated dependence on curvature and flow history [38]. Regardless of the specific correlation law and as a consequence of variations in the absolute value of the chemiluminescence intensity, the purpose of performing spontaneous luminosity mea- surements is to reveal differences in the normalized (by its maximum value for each flame) distri- bution of the chemiluminescence signal that can clearly point to flame structure differences among various fuels. Flame heightH fl is commonly used to compare flames in Bunsen and jet burner configurations. It is inversely proportional to the global turbulent flame speed,S T [13], and will be the main metric to compare between fuels. H fl can be defined from the chemiluminescence intensity along the centerline. For each flame, CH* luminosity are taken using a filter centered at 434 nm with a bandwidth of 17 nm to capture the primary emittance band from CH* which peaks around 431 nm. H fl is defined as the location where the line-of-sight chemiluminescence of the flame brush drops to a quarter of the maximum value along the jet centerline, not accounting for the luminosity from the pilot as seen in Fig. 2.4. The maximum intensity occurs in the pilot flame since the pilot is at stoichiometric conditions and has the highest reactivity and heat release. The error bars on these 22 Flame Length Location of Maximum Intensity Axial Distance (mm) Absolute Intensity Location of 25% Maximum Intensity Figure 2.4: Graphic showingH fl determination from the CH* Luminosity Profiles measurements are +/-1D. A 5% change in the threshold only leads to marginal changes in flame height and would affect all flames similarly. Thus, the observed trends will not be affected. 2.3 Particle Image Velocimetry Particle Image Velocimetry (PIV) is a technique that has been used for many years to characterize 2D and 3D flow fields and is the primary technique used in the current laboratory to measure laminar flame speeds of counterflow flames i.e. [15, 18]. The basic concept of the experimental technique is fairly straighforward. Tracer particles are added into a flowfield which are small enough to follow the flowfield. A laser sheet is expanded into a thin sheet to illuminate the tracer particles in a volume of interest. Two laser pulses at a discrete time interval are fired in order to determine the velocity field by correlating the movement of the particle field between the two images taken by a camera. The advantage of the PIV measurement technique is that it is a non-intrusive technique (un- like probe or hotwire measurments) that allows for accurate resolution of the flow field in time and 2D/3D space. This allows for the visualization of vortical structures that may be interacting with the flame front and the instanteous wrinkled appearance of the flame. As the jet flame is an axisymmetric configuration, 2D measurements along the centerline were taken to measure typical axial and radial velocity profiles on the left and right sides of the jet. Comparisons between the left and right side of the burner can be used to quantify asymmetries present in the jet. Care must be taken in order to minimize the uncertainty associated with this technique which include: particle drop out from the laser sheet due to beam misalignment, 3D effects of averaging of velocities over the beam thickness, peak locking errors which occur when the particles are on the order of a pixel, particles not adequately following the flow, and errors due to low particle density or lack of seeding. To remedy these errors, the laser beam was carefully aligned and the thickness and overlap of the beams was measured using burn paper. Also, a couple test images were taken 23 before each data set to ensure adequate particle seeding and correlation. The laser power was tested to make sure that the smallest diameter particles could be used while still creating visible texture and movement in the flow. Two separate lasers were used for this investigation. The first was a high power Q-switched frequency-doubled laser (Quantel Brilliant Twins) which generates light at 532 nm and expands using a plano-cylindrical lens (Thorlabs) into a thin sheet through the jet centerline at 10 Hz. The second laser was a Photonics DM-30 ND-YLF laser which produced light at 527 nm and was capable of investigating velocities of from a single shot to 10 kHz for time-resolved PIV . Unfortunately, the time resolution is not enough to resolve the smallest scales of the high Reynolds number jet flame. Care was taken to ensure that the beam profile was uniform between the two beams since a difference in laser intensity has been seen to skew the velocity field texture correlation. 2.3.1 Types of Particles and Particle Generation A variety of both solid and liquid tracer particles have been tested to ensure the most accurate flow tracers were used for the setup. First, for the unburned vectors, silicon oil has often been used in counterflow flames in the current laboratory [15, 18] since it evaporates around 500 K. This silicon oil has been estimated to have a diameter of approximately 0.3 microns when generated by high efficiency Laskin nozzles. A Laskin nozzle was used to generate silicon oil particles for the current setup. Using silicon oil with high velocities, however, presented an experimental challenge since silicon oil would build up over time as it hit the jet tube wall coating the inner jet tube and start to propagate upward due to the high velocity forcing against the wall creating buildup on the nozzle. For this reason, silicon oil was only attempted for a few lower velocity cases. Ethylene glycol which is notably more viscous was also tested, however, the signal intensity was too low for this liquid. Compared to liquid particles, solid particles have the advantage of surviving through the flame surface. Solid alumina particles of varying sizes were tested and injected into the system using a homemade bubbling seeder complete with a bypass. Since the scattering cross-section of light off of a particle is proportional to (d p =) 4 [41], light intensity increases exponentially with increase in particle diameter. Therefore, particle agglomerations can cause errors in correlation and unevenly bias the velocity vectors toward larger particles. To remedy this, particles of a slightly larger diameter were mixed with the predominantly smaller particles. This seeding system was heated to provide a higher flowrate through the chamber and remove moisture. Even with these precautions, a variety of particles was tested and agglomeration was still a problem until a new de-agglomerated alumina powder was found which produced far fewer large particles in the flow. The diameter of these particles was 0.05 to 0.3 microns. The smallest particles are preferred, however, depending on the viewing window and laser intensity, the larger 0.3 micron particles were needed for higher 24 signal to noise ratios. Solid particles are not perfect flow tracers, however. The Stokes drag equation allows for approximation of particle motion compared to velocity of mean flow. In the limit of large density difference as with alumina particles and air, the equation can be much simplified. In turbulent flows, the particle motion can be defined by a characteristic frequencyC [42]: C = 18 p d 2 p = 18! Sk 2 s (2.3.1) where s is the density ratio of the particle to fluid p = f , ! c is the highest frequency of turbu- lent motion, d p is particle diameter, and Sk is the Stokes number which is a characteristic non- dimensional frequency of particle response: Sk = ( ! c ) 1=2 d p (2.3.2) In the limit ofs>> 1, the particle velocity can be calculated by the following equation: u 2 p u 2 f = (1 + ! c C ) 1 (2.3.3) Assuming a maximum turblent frequency of interest of 10 kHz, the minimum particle diameter for fidelty of the tracer particles for alumina particles was calculated to be 0.78 microns [41]:. Therefore, having tracer particles of 0.05-0.3 microns as in the current experimental setup should be adequate to resolve the scales of turbulent motion. 2.3.2 Particle Image Displacement Errors andDt Optimization The PIV algorithm is dependent on particle displacement in a relatively short time period. Due to the finite thickness of the laser beam, if the time interval between laser pulses is too high, there may be particle dropout between images in the laser sheet. As the time intervals are on the order of 10s or less, and radial velocity is very small for an axisymmetric jet (on the order of 4% of the exit velocity), the risk of particle dropout is a non-issue for this configuration assuming a fi- nite laser beam thickness of 0.2-0.5 mm. An additional parameter of consideration is the particle displacement in pixels. A rule of thumb for good image correlation is to correlate the maximum velocity in the flow to around 0.25 times the size of the interrogation window box used. Neverthe- less, an optimization on the jet flame was conducted for variousDt which revealed that a particle displacement of around 4-10 pixels was ideal for a 16x16 window size. Larger particle displace- ments gave the same mean and r.m.s velocity fields. ADt optimization was performed for every new PIV configuration was performed in order to assure the best possible flowfield statistics. 25 2.3.3 Ignited Flame and Index of Refraction In addition to the other potential errors mentioned above, ignited flames present an even greater challenge to PIV methods. Stella et. al [43] conducted a thorough error analysis for PIV of flames. Two main challenges are: the thermophoretic force on the particle in the presence of the thermal gradient of the flame, and the uncertainty in the index of refraction due to temperature fluctuations along the path of laser light entering the camera after being scattered by the particles. The changes in index of refraction can causes particle position error for slow fluctuations resulting in systematic velocity error, fast fluctuations can lead to particle image blur [44]. These errors are found to be the most significant in areas of large velocity gradients. A thorough quantification of the errors from the fast fluctuations in refractive index from a propane gas flame show a systematic error of less than 5% of the exit velocity [44]. This is insignificant compared to the overall particle displace- ment. For the PIV method this error actually decreases with increasing velocity gradients. Stella et. al [43] also concluded that refractive index variations are negligible on a laboratory scale. How- ever, thermophoretic forces are present and can have a larger effect on velocity gradients causing a smoothing of the gradients at the flame surface. This force is proportional to particle diameter and temperature gradient and for small particles of 0.3 micron in size, the velocity lag measured in a flame was on the order of 0.3 m/s which is small in comparison to the velocities experienced by the jet flame. Therefore, these errors are neglected and added to the overall uncertainty value of 5% of the mean value. 2.3.4 Post-Processing Methods The image couples are post-processed with the Davis Flowmaster software (LaVision Research Inc.) with decreasing interrogation window sizes of 64x64 pixels then 16x16 pixels or even down to 8x8 pixels with 50 percent overlap between windows and then averaged to obtain the mean and r.m.s. velocity profiles. Vectors with a peak ratio of less than 1.05 are removed and the universal outlier detection method is used for spurious vector removal. Lastly, averaged and r.m.s. axial and radial velocity components were calculated. The algorithm for processing of the PIV vector fields was modified from the correlation image velocimetry (CIV) technique described by Fincham and Spedding [45]. The integral length scaleL 0 was estimated by integrating the longitudinal and transversal au- tocorrelation functions [6] ofu x andu r along the axial and radial directions with the formula: L 0 ( ! x ) (L xx ( ! x )L xr ( ! x ) +L rx ( ! x )L rr ( ! x )) 12 (2.3.4) 26 where L ij ( ! x ) = Z <V i ( ! x )V i ( ! x +z ^ j)> <V i ( ! x )V i ( ! x )> dz (2.3.5) with i and j indices corresponding to x and r. Velocity measurements with the highest spatial resolution are used to calculateL 0 . 2.3.5 Uncertainty Calculation The PIV uncertainty of the mean flow properties of the jet is dependent onU jet and is estimated to be0:05U jet . This is due to the fact that velocity vectors cannot be well resolved below a pixel displacement of 0.5 pixels. This displacement is held fixed at approximately 10 pixels as a rule of thumb for good correlation. This means that the optimum inter-frame time will change dependent on the exit velocity and will always be 5% of the maximum expected velocity at the jet exit. PIV measurements in the coflow and pilot have a 10% relative uncertainty due to a lower particle seeding density. 2.4 Temperature measurements Temperature measurements along the centerline will be taken with a ceramically-shielded ther- mocouple (Omega, Inc) and corrected for blackbody radiation. The thermocouple tip, 200m in diameter, is inserted into the flame perpendicular to the incoming flow with approximately 1 mm of the leads exposed. The thermocouple measurements have an uncertainty of50K everywhere except at the interface between flows. There, the uncertainties reach100K due to oscillations in the measured temperature values and the high shear present in the flow. 27 Chapter 3 Numerical Approach 3.1 PREMIX Knowledge ofS L of the reacting mixtures is necessary to determine the mixture ratios of the fuel/air mixtures investigated. S L was computed using the PREMIX code [46]. The PREMIX code cal- culates the species and temperature distribution in a freely propagating, isobaric, adiabatic 1D propagating flame front. The mass flow rate is used as an eigenvalue of the problem and requires a starting estimate of the location of the flame front, temperature, and reactant profile. The solu- tion is found by using a Newton solver and iterating toward a solution by using a time-integration algorithm until the final solution is reached. This process is very slow and the perturbations from time-stepping are small. Given the high activation energies present in the reaction layer, the prob- lem is stiff and has difficulty with convergence so a course mesh is used as a starting point and the solution is gradually refined to give a more accurate temperature, velocity and species profiles. It is modified to account for radiation at the optically thin limit and integrated with the CHEMKIN [47] thermal and the transport [48] subroutine libraries. For each flame, the flame thermal thickness, F , was computed in order to place the conditions considered in this study on a regime diagram. The flames were simulated using the detailed JetSurF 2.0 model for combustion of hydrocarbon fuels up to C 12 [49] 3.2 Senkin and Opposed Jet Code In an effort to understand how changing the temperature-time history of a reaction mixture affects heavy hydrocarbon reaction zones, adiabatic, isobaric partial reaction of n-C 12 H 26 /air mixtures were investigated first under atmospheric pressure using Sandias Senkin [50] and CHEMKIN [46]. The conditions considered include three equivalence ratios, = 0.7, 1.0, and 1.4, and four initial temperaturesT 0 = 1000 K, 1100 K, 1200 K, and 1300 K at which the fuel/air mixture was exposed to the high temperatures and allowed to decompose for set time intervals. The species distribution and composition 0-D partially reacted mixtures computed from Senkin was followed as a function of time, and used as input to laminar premixed flames calculations. The laminar flame speed and extinction strain rate K ext of partially reactant mixtures were computed using the PREMIX and opposed jet [51] codes, respectively. Defined as the maximum absolute value of the axial velocity 28 gradient in the hydrodynamic zone at extinction (e.g. [15]),K ext was computed in the twin-flame opposed jet configuration with a nozzle separation of 1.0 cm. The opposed jet code is based off of the OPPDIFF FORTRAN routine which computes the axisymmetric opposed jet solution. The chemical kinetics was modeled using JetSurF 1.0 [30]. The model is composed of 194 species and 1459 reactions and accounts for high-temperature oxidation kinetics of n-alkanes up ton-C 12 H 26 . JetSurF 1.0 is based on the H 2 /CO/C 1 -C 4 chemistry of USC-Mech II [31] and has been validated against flame propagation and extinction data [15]. 3.3 Large Eddy Simulations (LES) Recent Direct Numerical Simulations (DNS) with detailed chemistry [52, 53], have shown that the mean species and chemical source term profiles (in progress variable space) of high Karlovitz number flames can be well approximated by one-dimensional unstretched premixed flames. There- fore, the combustion modeling used in the present LES relies on chemistry tabulation of laminar flames with presumed sub-filter probablity density function(PDF). These simulations were per- formed at the California Institute of Technology as a collaboration with the experimental effort at the University of Southern California. 3.4 Turbulent Jet Flame Simulations Steady-state Reynolds-Averaged Navier Stokes (RANS) Simulations were performed of the tur- bulent jet flame in order to get a basic idea of how flame scalars interact with the mean and r.m.s velocity field, temperature and heat reelase field of the flame. Particularly, CH*, H 2 CO, OH, and HCO were of interest since the product of H 2 CO and OH is often used as a marker of heat release and CH* is used in this experiment as a marker of global reactivity. Additionally, as the 523 K temperature contour which marks the boiling point of silicon oil is used as a marker for the flame surface, the discrepency between the 523 K temperature contour and the CH* layer were of inter- est. The flame was modeled with the open source Computational Fluid Dynamics (CFD) solver OpenFOAM [54], in particular the solver reactingFOAM which is used for statistically steady-state combustion. Since the k-omega SST model captures the gradients of a jet flow better it was used instead of the typical k-epsilon model. Inlet conditions of k and! at the jet exit were approximated assuming a constant turbulence intensity of 10% at the jet exit using the equation:k = 3=2(IU jet ) 2 . ! the turbulence specific dissipation rate was estimated from: ! = k 0:5 C L 0 (3.4.1) 29 whereC =0.09 [55] andL 0 is the integral length scale of the flow. The combustion was modeled assuming a partially-stirred reaction model which defines a transient reaction rate dependent on a hybrid of the turbulent and chemical mixing time scales and assuming only a portion of the mixed fraction reacts. This requires an estimation of the coefficientC mix which is dependent on the turbulent Reynolds number For typical flowsC mix = 0:1. The equation forC mix is given by: C mix = s 1 1 +C Re t (3.4.2) These simulations were performed for a jet exit velocity of 56 m/s forn-dodecane/air flames using a 31 species skeletal mechanism forn-dodecane/air based off of JetSurF 1.0 [56]. A 2D axisym- metric domain was estabilished from the jet centerline up to amx=D of 40. The pilot and jet were simulated using hot ethylene/air combustion products at 1600K and an exit velocity of 4.5 m/s. The outflow conditions were set to freestream flow for velocity, temperature and pressure matching the conditions of the coflow and the species were set to zero gradient at the boundary. 30 Chapter 4 Effects ofn-Dodecane Decomposition on its Fundamental Flame Properties 4.1 Introduction It has been recognized for some time now that in premixed flames large hydrocarbon fuel molecules undergo endothermic decomposition in the preheat zone, transforming themselves into smaller molecular-weight fragments [57]. These small fragments include hydrogen, ethylene, methane, propene, the butane isomers and in some cases, one-ring or multi-ring aromatics. The heat trans- ported from the reaction zone to the preheat zone is partitioned into sensible enthalpy that heats the unburned fuel-air mixture and to some extent, the chemical enthalpy as the result of endothermic decomposition of the fuel [58]. The decomposed fragments, somewhat more energetic than the parent fuel, are then transported into the reaction zone and are oxidized, giving rise to bulk heat release for the flame. Studies of laminar flames of higher hydrocarbon fuels [15, 17, 59] have re- vealed that with the exception of benzene and toluene, all fuel molecules decompose readily in the preheat zone within the temperature range of around 1100 to 1450 K. For heavy n-alkane flames the time scale for fuel decomposition is less than 1 ms [17, 58]. Obviously changes in the structure of the preheat zone can impact the composition of the decomposition fragments, which in turn, influences the heat release and local burning rates. Turbulent flames tends to thicken the preheat zone [13, 24] where fuel decomposition occurs. The extent of thermal decomposition is dependent on the temporal temperature history and res- idence time of the preheat zone, and mixing to an extent. A longer residence time and higher temperature can lead to a greater degree of thermal decomposition, leading to the production of a greater amount of smaller, less hydrogenated fragments (e.g., ethylene, acetylene and hydrogen), or a greater preheat-zone endothermicity. Additionally, in turbulent flames, mixing time scales of the order of 1s-10 ms are typical under conditions of relevance to gas turbine combustors [60]. Under such conditions, mixing between the hot products and unburned reactants could take place at time scales comparable to those of fuel decomposition and thus it can promote fuel pyrolysis and modify the free-stream composition. Two factors can impact the local flame velocity: the production of smaller, less hydrogenated fragments can lead to a slower heating rate in the preheat zone thus reducing the flame propagation rate, but these fragments more energetic than the lesser 31 decomposed fragments (e.g., propene and 1-butene). The two effects are coupled and expected to cancel out each other to an extent, leading to a relative insensitivity of the flame response to ther- mal decomposition. The magnitude to which the coupling impacts the flame speed and extinction strain rates remains unknown. To date, our understanding of flame kinetics is achieved mostly through experimenting and modeling laminar flames in canonical experimental configurations (e.g., [61, 62]). Many studies have been centered on the characterization of flame properties such as laminar flame speeds,S L , and extinction strain rates, K ext , in order to test chemical kinetics models [[63]] and explore the underlying reaction pathways and kinetics. Accurate knowledge ofS L can be essential to an un- derstanding of premixed turbulent flames [64, 65]. Here, we shall address the question about the coupling of fuel thermal decomposition and heat release from the oxidation of the more energetic decomposition products by computational experiments using n-dodecane (n-C 12 H 26 ) combustion as an example. n-Dodecane was chosen because it is a representative component of jet fuels and its high-temperature kinetics is relatively well characterized (e.g., [17, 66–68]). The goal of the current investigation was to assess the effects of fuel decomposition on laminar flame propagation and extinction. 4.2 Computational Setup and Rationale In order to focus on the detailed effects of fuel pyrolysis in the reaction zone, a detailed kinetic model developed for high temperature n-alkane oxidation, the JetSurF 1.0 [69] kinetic model was used. JetSurF 1.0 is based on the H 2 /CO/C 1 -C 4 chemistry of USC-Mech II [70] and has been validated against flame propagation and extinction data [15] Using the Premix [46] code coupled with CHEMKIN II [47] and transport libraries [48], a laminar freely-propagating premixed n- C 12 H 26 flame was first simulated at stoichiometric conditions at an initial temperature (T 0 ) of 403 K which, based on the Clausius-Clapeyron relationship allows the full vaporization of up to =0.7-1.4n-C 12 H 26 /air mixtures. The flame structure shown in Figure 4.1 reveals that there is a distinct region of pyrolysis ofn-C 12 H 26 before the flame front before the fuel fragments of mostly n-alkenes and methane are oxidized in the flame front. The pyrolysis occurs due to heat transfer from the flame front within a temperature range of 1050-1450 K. However, the effect of thermal decomposition of the parent molecule on the structure of a freely propagating flame is difficult to isolate since the temperature-time history of the reactants varies rapidly before the fuel fragments are consumed in the flame front. Since the ideal laminar flame structure may not be preserved in practical applications, this study aims to decouple the thermal decomposition of the fuel from the reaction front. This allows an in-depth study on how changing the temperature-time history of the reactants before the flame surface can affect un-stretched and 32 Figure 4.1: Structures computed for atmospheric-pressure laminar premixed stoichiometric n- dodecane-air flames of unburned gas temperature of 403 K. stretched laminar flame properties. For example, perturbations near a turbulent flame front may cause instantaneous heating followed by quenching due to convection away from the flame front, affecting the breakdown of n-C 12 H 26 and the composition of fuel fragments entering the flame front. Experimentally, the most common technique to study thermal decomposition is to place n- C 12 H 26 in a flow reactor to examine breakdown of dilute fuel-N 2 mixtures at elevated initial tem- peratures and under adiabatic and isochoric conditions as pictured in Fig. 4.2. Following that approach, the computational experiments were performed in multiple steps to mimic a systematic experimental setup using a flow reactor for mixture composition and the opposed jet twin flame setup to extract both laminar flame speed,S L , and extinction strain rate,K ext . Figure 4.3 illustrates the computational approach. First, initial baseline mixtures of un-decomposed n-C 12 H 26 /air atT 0 = 403 K and = 0.7, 1.0, and 1.4 were defined as a starting point. These mix- tures were then heated instantaneously to three pyrolytic reactor temperatures (T p;0 = 1100 K, 1200 K, and 1300 K). T p,0 are chosen to be in the middle of the temperature window in which high temperature pyrolysis ofn-dodecane occurs as shown in Fig. 1. These mixtures were then allowed to react under adiabatic and isobaric conditions at atmospheric pressure using Senkin [71] coupled to CHEMKIN II [47]. The species distribution and composition of the partially reacted mixtures, 33 Figure 4.2: Time history computed for the thermal decomposition of 1.12% (mol)n-dodecane in air at 1 atm and initial temperature of 1400 K under adiabatic and isochoric conditions, plotted in logarithmic (left panel) and linear, zoomed-in (right panel) time scales. The horizontal lines on the right side of the left panel indicate the equilibrium composition of gas-phase species. X p; was followed as a function of time. The reaction was stopped at different and the mixtureX p; instantaneously cooled to a initial temperatureT 0 0 in order to match the enthalpy of the initial mixture and ensure identical adiabatic flame temperature Tad as the natural laminar remixed flame. These cooled mixtures of X p; at T 0 0 were then used as input to laminar premixed flames calculations and compared against the un- decomposedX 0 cases atT 0 .S L andK ext of the partially reacted mixtures were computed using the Premix [46] and OPPDIF [72, 73] codes, respectively. Defined as the maximum absolute value of the axial velocity gradient in the hydrodynamic zone at extinction (e.g., [18],K ext was computed in the twin-flame opposed jet configuration with a nozzle separation of 1.4 cm. 34 Figure 4.3: An illustrative diagram of the computational experiments performed. The dashed line encloses an overall adiabatic reactor comprised of four processes: 1) the fuel-air mixtureX 0 at an initial temperatureT 0 = 403 K and pressure of 1 atm is heated to the pyrolytic reactor temperature T p;0 , 2) the mixture undergoes adiabatic, isobaric thermal decomposition for a residence time of to obtain a partially reacted mixtureX p; , 3) cooling the partially reacted mixture to reachT 0 0 from T p;1 (to ensure identical adiabatic flame temperature T ad as the natural laminar premixed flame) after 4) an adiabatic laminar premixed flame. 4.3 Pyrolysis ofn-C 12 H 26 Initial reactions in the adiabatic and isobaric oxidation ofn-C 12 H 26 in air involve mainly fuel de- composition into H 2 , CH 4 , and small alkene fragments. Being endothermic in nature, fuel cracking results in an initial temperature drop as shown in Fig. 4.4, followed by a temperature rise due to the oxidation of the partially reacted mixture. In all cases, the sensible energy is transformed par- tially into chemical enthalpy initially due to endothermic decomposition. As expected, with the initial temperatureT p;0 increased from 1100 K to 1300 K, fuel decomposition accelerates. These behaviors have been observed commonly in shock tube and flow reactor studies [74–76]. The composition variations of the intermediates during the initial stage of adiabatic-isobaric n-dodecane oxidation were computed for three T p;0 values of Fig. 4.4. All species with mass percentage> 0.01% of the initialn-C 12 H 26 were considered as input to computingS L andK ext . Representative mass percentage values of the key intermediates are shown in Table 1 for = 1.0 andT p;0 = 1200 K. In agreement with previous findings [17], the most abundant intermediates formed during the partial reaction of then-C 12 H 26 /air mixture are H 2 , CH 4 , and C 2 -C 4 alkenes, with C 2 H 4 being the most dominant. Larger alkenes (C 5 -C 9 ), as well as H 2 O, CH 2 O, and CO are present also but they 35 Figure 4.4: Temperature (top), and mass fraction-time history (bottom) computed for adiabatic and isobatic oxidation of n-dodecane-air mixtures at three equivalence ratios () with initial tempera- ture of a)T p;0 = 1100 K, b)T p;0 = 1200 K, and c)T p;0 = 1300 K, all at 1 bar. are in smaller quantities. Lower initial temperatures result in greater amounts of larger alkenes and oxygenated compounds, while higher temperatures produce greater amounts of H 2 and C 2 H 4 . In order to illustrate the effect of cracking temperature on the distribution of hydrocarbon frag- ments formed, the mass fractions were weighted by the initial mass fraction ofn-C 12 H 26 and the results are shown in Fig. 4.5 for = 1.0; nearly identical results were found for = 0.7 and = 1.4 mixtures. H 2 , CH 4 , and n-alkenes with less than five carbon atoms, are generated in larger amounts with increasing n-C 12 H 26 decomposition. Fig. 4.5(a) and (b) show that production of C 2 H 4 and H 2 increases with temperature as opposed to CH 4 that appears in smaller amounts at 1300 K. The percentages of C 3 and C 4 alkenes do not depend onT p;0 but only on the amount of n-C 12 H 26 remaining. The mass fractions of the larger alkenes first increase and then decrease as they begin to decompose themselves into smaller fragments, and representative results are shown in Fig. 4.5(a). Finally, the mass fractions of the oxygenated compounds increase gradually and subsequently more rapidly with time andn-C 12 H 26 consumption, as shown in Fig. 4.5(c). Analysis of the computed heat release and reaction rates revealed the controlling mechanisms of the observed fuel decomposition behavior. In the initial stages, the C-C fission inn-C 12 H 26 is responsible for forming various alkyl radicals. -Scission of the alkyl radicals produces smaller 36 Figure 4.5: (a-c) Mass percentages of key products computed forn-C 12 H 26 decomposition in air at = 1.0 and several initial temperaturesT p;0 , and (d) net production rates of selected species as a function of reaction time for = 1.0 andT p;0 = 1200 K. fragments sequentially and notably, C 2 H 4 throughn-C 3 H 7 ! C 2 H 4 + CH 3 , C 2 H 5 +M! C 2 H 4 + H +M, or pC 4 H 9 ! C 2 H 4 + C 2 H 5 . As the reaction progresses, H-abstraction from the fuel by the H and CH 3 radicals becomes more prevalent. A buildup of radicals ensues, leading to exothermic oxidation reactions of the partially reacted mixture that produces mainly H 2 O, CH 2 O, HO 2 , and CO, and eventually temperature rise. Species net reaction rates were computed as a function of the time. The results are shown in Fig. 4.5(d) forT p;0 = 1200 K and = 1.0 and at different percentages ofn-C 12 H 26 decomposition. The destruction rate ofn-C 12 H 26 reaches a maximum at a very early stage, and this maximum is reached progressively earlier with increasingT p;0 . At that particular time, the production rate of C 2 H 4 (and similarly C 3 H 6 ) reaches also a maximum and H 2 , CH 4 , as well as other alkenes are being generated at the same time. At a later time, the net production rates of the heavier alkenes 37 (e.g., 1-C 6 H 12 ) change sign as their consumption rates become larger than the production rates. This is followed by the rapid production of H 2 O and CO. Table 4.1 shows the compositions of the partially reacted mixturesX p;r at different degrees of fuel decomposition used to computeS L andK ext . Table 4.1: Species and their mass fractions of during the initial stage ofn-dodecane oxidation in air over a range of time or %n-dodecane decompossed. The computation was made for = 1.0 and several initial temperatureT p;0 . 4.4 Propagation of mixtures or air with partially decomposed n-C 12 H 26 In order to assess the effects of fuel decomposition on flame propagation and extinction, compar- isons were made by keeping the total enthalpy of the unburned mixture constant, which allows for the evaluation of the effects of energy partition between the chemical and sensible enthalpies. For each, baseline mixturesX 0 were defined as un-decomposedn-C 12 H 26 /air flames at an unburned mixture temperatureT 0 = 403 K similar to a previous study [10]. As a reminder, the initial tem- perature valuesT 0 0 for the partially reacted mixtures were adjusted for eachX p; to match its total 38 enthalpy to that of the un-decomposed mixture such thatT ad , ofX p; matchedT ad , atX 0 . For the partially reacted mixtures, values of T 0 < 403 K were used reducing thus the sensible enthalpy in order to compensate for the increased chemical energy stored into the products of fuel decom- position. In order to verify that the T 0 0 values would result in constant enthalpy, the Equil code was used to computeT ad , of eachX p; . T ad was verified to remain constant for each regardless of the extent of fuel decomposition. The resulting conditions forT 0 0 and times are shown also in Table 4.2. Table 4.2: Unburned mixture conditions used in flame simulations. Given that the flame simulations were performed at different T 0 0 , the comparisons between the various cases were made by using as observables the mass burning rate _ m u = u S L and the density-weighted extinction strain rate u K ext , where u is the density of the unburned mixture. Figure 4.6(a) depicts the percent variation of _ m u for decomposedn-dodecane/air mixtures from the un-decomposed mixture (the initial reference mixture of n-dodecane/air at X 0 = 403 K and different ) for various mixtures obtained from the pyrolysis reactor atT p;0 = 1100 K, 1200 K, and 1300 K. The largest difference is about 15% for the = 1.4 case, indicating that flame propagation is not very sensitive to the partial reaction of the fuel-air mixture. Nevertheless, _ m u increases with the extent of n-C 12 H 26 decomposition. As mentioned earlier, a greater extent of decomposition 39 Figure 4.6: (a) Ratio of the mass burning rate _ m u of a decomposedn-dodecane/air mixture at three representative initial temperature T p;0 values to that of un-decomposed mixture as a function of percentn-C 12 H 26 decomposed at = 0.7, 1.0, and 1.4. (b) Correlation between the mass burning rate enhancement and the ethylene yield in mass fraction. leads to an increase in production of H 2 and C 2 H 4 , both of which contribute to higher S L and hence higher _ m u . The rate of change in _ m u with the extent ofn-C 12 H 26 decomposition increases with as shown in Fig. 4.6(a). The variation of _ m u with the extent ofn-C 12 H 26 decomposition for the = 1.0 and = 1.4 cases was determined to be relatively insensitive toT p;0 . Figure 4.6(b) depicts the variation of _ m u with the mass fraction of C 2 H 4 for all. The percent difference in _ m u varies almost linearly with mass fraction of C 2 H 4 and is independent of. The logarithmic sensitivities of _ m u to rate parameters and binary diffusion coefficients were computed in Fig. 4.7. For all , _ m u is sensitive to mostly H + O 2 ! O + OH and CO + OH ! CO 2 + H reactions, and the O 2 -N 2 binary diffusion coefficient (e.g.,[15, 61]). Given the small change in _ m u , it is not surprising that the sensitivities of _ m u to kinetics and binary diffusion coefficients do not change much with the extent of fuel decomposition, as shown in Fig 4.7 for =1.0. A minor effect can be seen for the sensitivity to CO + OH! CO 2 + H that increases with the extent of fuel decomposition. As reported in previous studies (e.g., [3]), _ m u is sensitive to the kinetics of reactions involving H 2 , CO, and C 1 -C 3 hydrocarbons. Additionally the sensitivity to the O 2 -N 2 binary diffusion coefficient decreases slightly with the extent of fuel decomposition. 4.5 Extinction of crackedn-C 12 H 26 /air flames The density-weightedK ext ( u K ext ) was examined in a manner similar to _ m u . The results are sum- marized in Fig. 4.8. In contrast to _ m u , fuel decomposition can cause u K ext to change considerably 40 Figure 4.7: Ranked logarithmic sensitivity coefficients of the mass burning rate _ m u with respect to (a) rate parameters and (b) binary diffusion coefficients for three reactant mixture conditions. from the un-decomposed value. For smaller extents ofn-C 12 H 26 decomposition, the deviation from the un-decomposed is small, as expected. For larger extents ofn-C 12 H 26 decomposition though, u K ext can vary compared to the un-decomposed case by as much as 170% for = 0.7, to less than 15% for = 1.4, for allT p;0 values considered. Due to the high molecular weight ofn-C 12 H 26 , an-C 12 H 26 /air mixture have the Lewis numberLe > 1 for < 1.0 andLe < 1 for > 1.0. For < 1.0 stagnation flames that are positively stretched, the overall flame reactivity decreases with stretch while the opposite is true for> 1.0 flames (e.g., [77]). For 1.0 flames withLe> 1 u K ext increases linearly with the extent of fuel decomposition. This is a result of the decrease of Le as the extent of fuel decomposition increases and lighter species are present in the unburned mixture. AsLe decreases, the overall reactivity increases with stretch and as a result the resistance to extinction increases as well. For> 1.0 flames withLe< 1 on the other hand, increasing the percentage of light species in the unburned mixture as the extent 41 Figure 4.8: Ratio of u K ext of decomposedn-dodecane-air mixtures at several representative inital temperature T p;0 values to that of un-decomposed reactant mixture (top panels) and ranked log- arithmic sensitivity coefficients of u K ext with respect to binary diffusion coefficients for three reactant mixture conditions. of fuel decomposition increases does not have any notable effect on the mixturesLe, and as a result the extinction propensity is minimally affected. The logarithmic sensitivity coefficients ofK ext to rate parameters and binary diffusion coefficients were computed. Similarly to _ m u ,K ext is sensitive largely to the H 2 , CO and C 1 -C 3 hydrocarbon kinetics for all extents ofn-C 12 H 26 decomposition so the reader is referred to Fig. 4.7 for a similar analysis. It is clear from this analysis that the up to 42 150% differences in u K ext t cannot be attributed to differences in kinetics resulting from unburned composition. The sensitivity of K ext to binary diffusion coefficients seems to provide a more interesting explanation for the differences in u K ext . The results were very similar for allT p;0 values, and only the results forT p;0 = 1100 K are shown in Fig. 4.8. For the = 1.4 case, the sensitivity to the O 2 -N 2 binary diffusion coefficient is by far the highest because being the deficient reactant, the O 2 transport rate is critical to the overall reaction rate. Thus, the modification in the fuel composition due ton-C 12 H 26 decomposition has little to no effect onK ext . For the = 0.7 and = 1.0 cases, the results of Fig. 4.8 indicate that K ext is mainly sen- sitive ton-C 12 H 26 -N 2 and C 2 H 4 -N 2 binary diffusivity coefficients, and that the magnitude of the sensitivity depends on the fuel composition in the unburned mixture. These results are reasonable since the main product of the fuel decomposition is C 2 H 4 . Thus,K ext becomes more sensitive to C 2 H 4 diffusion as its concentration in the unburned mixture increases with the extent ofn-C 12 H 26 decomposition. Since C 2 H 4 is more diffusive thann-C 12 H 26 , its larger diffusion coefficient causes K ext to increase. By extension, the propagation speed of stretched and/or wrinkled laminar flames also exhibit high sensitivity to fuel decomposition because of the preferential diffusion of the small and highly reactive decomposition products. Based on the computed stretched flame structures, the mass- burning rate at extinction was derived and it was shown that it can exceed the laminar mass-burning rate of the un-decomposed case by as much as 40-50% for = 0.7 flames due to the synergistic effects of stretch andLe. This behavior has significant implications for the determination of the turbulent flame speed since turbulent flames are highly stretched and wrinkled. 4.6 Concluding Remarks Under intense turbulent conditions encountered in high-speed air breathing propulsion applica- tions, large molecular weight fuels are expected to undergo considerable decomposition in both the free-stream as well as within a thickened preheat zone. Under such conditions, the energy transported from the flame into the unreacted mixture can be partitioned to various extents be- tween chemical and sensible. The adiabatic, isobaric, partial reaction ofn-dodecane/air mixtures was modeled under atmo- spheric pressure at several initial temperatures and equivalence ratios. It was shown that fuel de- composition precedes the oxidation of decomposed intermediates. The decomposition occurs over times scales of less than 5 ms, which is of the same order of the fluid mechanics timescales encoun- tered in high Reynolds number turbulent flows. Results indicate that the n dodecane decomposes mainly to form hydrogen, methane, C 2 -C 4 alkenes, and some larger alkenes and is mainly sensi- 43 tive to the reactivity and diffusivity of these smaller hydrocarbon fragments. The mass burning rates and density-weighted extinction strain rates of laminar premixed flames of the partially re- acted mixtures were computed and compared to those of unreactedn-dodecane/air mixtures. In all cases, the total enthalpies of the unburned mixtures were kept constant so as to provide meaningful comparisons. The results show that the impact of fuel decomposition on the laminar mass-burning rate is finite but relatively small, that is at most 15% compared to the un-decomposed case ofn- dodecane/air mixtures for the conditions considered herein. Furthermore, the mass-burning rate was determined to scale linearly with the mass fraction of ethylene in the mixture, indicating that the more reactive and diffusive species plays a role in modifying the laminar flame speed. The density-weighted extinction strain rate was found to be notably more sensitive to fuel decompo- sition under fuel-lean and stoichiometric conditions due to the effect of Lewis number and differ- ential diffusion effects on stretch. However, the extinction behavior of fuel-rich flames is rather insensitive to the extent of fuel decomposition. Sensitivity analysis showed that the increased re- sistance to extinction as the extent of fuel decomposition increases is not caused by kinetics but rather by the increased diffusivity of the lighter species such as for example ethylene. These results are important in defining proper scaling parameters in turbulent combustion stud- ies as well as in developing kinetic models for heavy practical fuels. Regarding the scaling pa- rameters, it is essential to assess whether the laminar flame speed based on the parent hydrocarbon molecule is an appropriate scaling parameter given that the degree of fuel decomposition has an a relatively small but still finite effect on the laminar flame speed and a notable effect on flame ex- tinction and stretched flame propagation. Regarding the development of kinetic models, the results of the present analysis suggest that it is possible to decouple the fuel and the lighter products of fuel decomposition kinetics, with the latter controlling high-temperature flame phenomena. Fur- ther investigations are needed to address these two important issues that are key towards providing quantitative predictions of real combustors. Particularly of interest is the increase in ethylene and identifies its role as an important intermediate. It will be studied experimentally along with the heavy hydrocarbon fuels and in addition to methane. 44 Chapter 5 Heat Loss and Fuel Effects on Global Characteristics of Piloted Premixed Jet Burner 5.1 Introduction Though turbulent premixed combustion has been studied widely, few numerical and experimental studies have assessed finite-rate chemistry or fuel effects, especially at high turbulence levels. Fig- ure 5.1 depicts a Borghi/Peters diagram [9, 10] with a representative survey of nearly 140 turbulent premixed flame experiments. Two main observations may be drawn from this survey. First, the vast majority of the experimental investigations have been carried out for either hydrogen or methane flames, with very limited data on heavy hydrocarbons despite their omnipresence in practical com- bustion engines [78, 79]. Second, there is a scarcity of data for any fuels (and especially for high carbon number fuels) in the thin and broken reaction zone regimes. One of the overarching goals of the present paper is to provide much needed data for heavy hydrocarbons at highKa in the thin and broken reaction zone regimes on the Borghi/Peters diagram. Independent of the lack of experimental data for heavy hydrocarbon fuels under high turbu- lent intensities, such Borghi/Peters diagram highlights one important underlying assumption with respect to the modeling of turbulent premixed flames. The use of only two laminar flame proper- ties, namely the laminar flame speed,S L and the laminar flame thickness, F , makes the implicit assumption of no significant chemical effects beyond those global properties. Stated differently, the structure of a turbulent flame is assumed to be the same as that of the one-dimensional freely propagating flame (decribed in detail in the background chapter). This assumption is used in many turbulent combustion models [80–82]. This raises an important question: what is the impact of the fuel and its chemistry on the turbulence-flame interaction? Previous numerical simulations at high Karlovitz numbers using detailed chemistry focused mainly on simple fuels such as hydrogen [83, 84] and methane [85, 86]. Only a few studies considered heavier hydrocarbon fuels such asn-heptane. These studies have shown persistent dif- ferential diffusion effects inn-heptane flames at the transition from the thin to broken/distributed reaction zone regimes [52, 53]. Following these observations, the goal of the paper is to investigate 45 Figure 5.1: Borghi diagram showing experimentally investigated fuels. Present work is shown indicated by red x’s. fuel and hydrodynamic effects at high Karlovitz numbers. Specifically, the objectives are two-fold. The first objective is to generate well-characterized experimental data for a variety of fuels rang- ing from small alkanes/alkenes (methane/ethylene) to large n-alkanes (propane and n-heptane) to aromatic fuels (toluene) under extremely lean conditions. The second objective is to identify pos- sible fuels effects under a fixedS L . These objectives will be achieved by using a modified Piloted Premixed Jet Burner (PPJB) [27, 28] combined with matching Large Eddy Simulations (LES) per- formed by the California Institute of Technology. Fuel effects will be isolated by conducting the experiments with fuel/air mixtures with the exact sameS L and by performing LES with models purposely assuming the same structure as laminar flames. First, an overview of the experimental setup is provided, followed by a description of the mea- surements techniques and the modeling strategies used to simulate the turbulent flames. Then, a detailed comparison of boundary conditions and jet development between experiments and simu- lations is provided. Finally, the effects of fuel and fluid dynamics are analyzed by considering the height of the turbulent flame. 46 Table 5.1: Inlet Conditions and Specifications of Parametric Heated PPJB Investigation 5.2 Experimental and Numerical Approach The PPJB developed at USC as discussed in previous studies [40, 87], also described in Chapter 2 was the basis of the experimental investigation. In order to fully characterize the burner, several parameters were varied. The three-stream configuration consisted of a central jet, a surrounding pilot flame to anchor the central jet to the burner, and a large coflow surrounding the pilot. As in the previous studies, the pilot was kept constant with a low velocity stoichiometric methane/air flame per the conditions listed in Table 5.1. , andU jet were varied to allow investigation of a range of laminar flame speeds which were outside the limits of the previous investigation with coflowing air. The coflow was ignited at varying temperatures using a mixture of lean H 2 and C 2 H 4 and air. The coflow can safely operate at a maximum temperature of 1850 K. Experiments and simulations are performed at various unburnt jet Reynolds numbers: Re jet =U jet D= = 25,000, 37,500, 50,000, and 75,000 for five different fuels: methane, ethylene, propane,n-heptane, and toluene. The experimental investigations were two-fold: the first condition matrix investigated heated coflow withT coflow = T ad,jet . The second investigated the effect of heat loss on the stability characteristics of the central jet flame by varying fixedT coflow temperatures ofT coflow = 1400 K, 1500 K, and 1600 K. The temperature of the coflow was measured using a B-type thermocouple to ensure consistency between data sets. 47 The unburnt mixture temperature was 298 K in all cases and the equivalence ratio of the main jet was varied to keep the same laminar flame speed,S L =11.3 cm/s. The methane flame has=0.60 with an adiabatic flame temperature T ad =1666K; ethylene =0.48, T ad =1543K; propane =0.56, T ad =1623K;n-heptane=0.56,T ad =1629K; and toluene=0.58,T ad =1720K. Table 5.2 provides a summary of the conditions investigated. The laminar flame thickness of methane/air was used in Table 5.2 because less than 7% variation was observed between the different fuels. These turbulent premixed flames were characterized by large turbulent intensities and high Karlovitz numbers. The flames were expected to fall either in the thin reaction zone regime or in the broken reaction zone regime. U jet (m/s) Re jet u 0 S L u 0 (m/s) t = u 0 L int (ms) L int f L 0 (mm) = q L intu u 03 (ms) Ka= f Da= t f Re t = u 0 L int u 68.5 25000 65 8 0.88 7.1 6.6 0.0153 524 0.0314 3037 100 37500 104 12 0.48 6.7 6.2 0.0071 1132 0.0173 4564 133 50000 157 18 0.32 6.3 5.8 0.0039 2079 0.0115 6405 200 75000 252 29 0.18 5.5 5.1 0.0018 4533 0.0063 9074 Table 5.2: Conditions of the experiments and simulations performed. Re t is the turbulent Reynolds number,Ka is the Karlovitz number, andu 0 is the peak turbulent intensity atx=D=15. The integral length scaleL int is evaluated in the shear layer atx=D = 15. All values are calculated based on the kinematic viscosity of the unburnt mixture but the turbulent properties of the flow, i.e. L int andu 0 used in Re t are measured with the flame ignited. S L and f , were calculated using PREMIX [46]. f was calculated for all fuels, but since less than 7% variation in value exists between the fuels only the value for methane was used in the table. 5.2.1 CH* Chemiluminescence Flame height H fl is commonly used to compare flames in Bunsen and jet burner configurations. Therefore,H fl was the main metric used to compare between fuels. For each flame, 100 images of the CH* luminosity were taken using a filter that was centered at 434 nm with a bandwidth of 17 nm to capture the primary emittance band from CH* that peaks around 431 nm. These images were taken with an Andor Zyla camera using an exposure time of 200 ms and then averaged together. H fl is defined as the location where the line-of-sight chemiluminescence of the flame brush drops to a quarter of the maximum value along the jet centerline, not accounting for the luminosity from the pilot. The error bars on these measurements wer e calculated to be +/-1D. 48 5.2.2 Particle Image Velocimetry (PIV) A high power Q-switched frequency-doubled laser (Quantel Brilliant Twins) generates light at 532 nm and expands using a plano-cylindrical lens (Thorlabs) into a thin sheet through the jet centerline at 10 Hz. The jet is seeded with 0.3-1m aluminum oxide particles that survive the flame. 500 statistically independent Mie scattering image couples are taken with three stitched high resolution SCMOS cameras (Andor Zyla 5.5) focalized with a F85mm/2.8D Nikkon lens in order to find the velocity field along the entire height of the jet. A laser-line filter at 532 nm removes the flame luminosity signal. Inter-frame times range between 1s-20s, depending onU jet and the average velocity at the location of interest. The image couples have a resolution of 57.9 m/pixel and are post-processed with the Davis Flowmaster software (LaVision Research Inc.) with decreasing interrogation window sizes of 64x64 pixels then 16x16 pixels with 50 percent overlap between windows and then averaged to obtain the mean and r.m.s. velocity profiles. Vectors with a peak ratio of less than 1.05 were removed and the universal outlier detection method was used to remove spurious vectors. PIV measurements at 1 mm above the burner surface (boundary conditions) are taken with a finer resolution of approximately 9m/pixel, using one SCMOS camera and a F200mm/4D Nikon lens. All three streams are seeded simultaneously with tracer particles. A separate t is used for the central jet and the pilot/coflow beacause the exit velocity of the central jet was significantly greater than the pilot and coflow streams. For this data set, 1000 statistically independent images are averaged. The PIV uncertainty of the jet is dependent on U jet and is estimated to be0:05U jet . PIV measurements in the coflow and pilot have a 10% relative uncertainty due to a lower particle seeding density. 5.2.3 Temperature measurements Temperature measurements along the jet exit plane are taken with a ceramically-shielded thermo- couple (Omega, Inc) and corrected for radiation. The thermocouple tip, 200m in diameter, is inserted into the flame perpendicular to the incoming flow with approximately 1 mm of the leads exposed. The thermocouple measurements have an uncertainty of50K everywhere except at the interface between flows. There, the uncertainties reach100K due to oscillations in the measured temperature values. 49 0 1 2 3 4 5 6 7 8 0 500 1,000 1,500 2,000 r/D T(K) Thermocouple LES Figure 5.2: Comparison of experimental and numerical temperature profiles atx = 1 mm above the jet exit. 5.3 Experimental/numerical validation Boundary conditions at the jet exit were examined first to ensure consistency between experiments and simulations. They are followed by a comparison of the downstream development of the turbu- lent jet. This is done for the methane/air flame at Re jet =50,000 andT coflow =1500K. Previous studies in similar PPJB [27, 28] have reported velocity profiles and species/temperature measurements at locations no closer thanx=D=5 andx=D=2.5, respectively. 5.3.1 Boundary Conditions As shown in Fig. 5.2, thermocouple measurements indicate that the temperature at the burner exit on the centerline stays close to the unburned mixture temperature of 298 K even with the pilot and coflow ignited. The temperate rises sharply in the pilot and peaks around T ad of stoichiometric methane/air at 2250 K. The temperature then lowers down to reach a constant value of 1500K in the coflow. The numerical and experimental temperature profiles match reasonably well. The dip in temperature between the pilot and the coflow (missing in the simulations) can be attributed to conduction heat transfer from the hot coflow to the pilot nozzle which was not accounted for in the simulations. Next, the velocity profiles at the jet exit (1 mm above the exit) are compared in Fig. 5.3. The measured velocities on the left and the right sides of the burner appear to be reasonably symmetric and consistent. The numerical simulations present a flatter profile with a sharper drop-off, suggest- ing that the experimental pipe flow is not fully turbulent (Fig. 5.3a). The difference in jet velocity 50 0 0.1 0.2 0.3 0.4 0.5 0 20 40 60 80 100 120 140 160 0 0.1 0.2 0.3 0.4 0.5 0 10 20 30 40 r/D ¯ U (m/s) u 0 (m/s) ¯ U Left ¯ U Right ¯ U LES u 0 Left u 0 Right u 0 LES (a) Jet - Axial velocity 0.5 1 1.5 2 2.5 3 0 2 4 6 8 10 r/D ¯ U (m/s) Left Right LES (b) Pilot/coflow - Axial velocity Figure 5.3: Comparison of the velocity profiles measured experimentally and predicted numeri- cally at x=1 mm above the jet exit. The vertical dashed line (right) indicates the edge of the pilot. profiles results in different shear layers close to the burner exit. This is consistent with the discrep- ancies in r.m.s. velocity (u 0 ) and entrainment of the pilot (Fig. 5.3b). However, these differences in inlet boundary conditions do not significantly affect the downstream development of the jet, as shown in Section 5.2. The comparison between experimental and numerical boundary conditions display sufficient agreement to give confidence that the simulations reproduce most of the geometry and burner- dependent effects on the exit plane. Further analysis is required to see if the behavior of the jet can be reproduced downstream in the jet evolution as the statistical properties and mean velocity profiles of the jet are responsible for the stabilization point of the flame which ultimately influencesH fl . 5.3.2 Jet development After confirming alignment between experimental and numerical boundary conditions, the next step is to investigate the downstream evolution of the turbulent reacting jet. This is done first by comparing mean ( U) and r.m.s. velocity (u 0 ) profiles along the jet centerline. Figure 5.4a shows a fairly good agreement between experiment and simulation as the centerline velocity shows the same general decay in mean velocity with a corresponding increase in r.m.s. velocity. This is typical of momentum spreading radially outward as the axial distance increases. The radial profiles of mean and r.m.s. velocity upstream show similar jet spreading behavior as seen in Figs. 5.4b and 5.4c. The LES mean radial profiles are slightly wider with higher r.m.s. velocity, a behavior which remains consistent further downstream. 51 0 10 20 30 40 50 0 0.2 0.4 0.6 0.8 1 1.2 0 10 20 30 40 50 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 ¯ U/ ¯ U jet x/D u 0 / ¯ U jet ¯ U PIV ¯ U LES u 0 PIV u 0 LES (a) Centerline 0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 0 0.05 0.1 0.15 0.2 0.25 r/D ¯ U/ ¯ U jet u 0 / ¯ U jet ¯ U Left ¯ U Right ¯ U LES u 0 Left u 0 Right u 0 LES (b) x=D = 15 0 0.5 1 1.5 2 2.5 3 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 0 0.05 0.1 0.15 0.2 0.25 r/D ¯ U/ ¯ U jet u 0 / ¯ U jet ¯ U Left ¯ U Right ¯ U LES u 0 Left u 0 Right u 0 LES (c) x=D = 25 Figure 5.4: Axial (left) and radial (center and right) profiles of mean U and r.m.s. velocitiesu 0 . 5.4 Global fuel effects The stability of flames in the original PPJB configuration [28] was observed to be influenced by, Re jet , andT coflow . In this study, is first adjusted to force a constantS L as baseline for comparison between fuels. While a single global quantity,H fl encompasses both hydrodynamic and chemical effects. H fl is defined from the centerline luminosity profiles which will be discussed next. is then varied for various fuels withT coflow =T ad,jet to studyH fl as an observable. Re jet andT coflow are then varied to investigate possible fuel effects on the flame height,H fl . 5.4.1 Global Flame Shape and Centerline Luminosity Profiles Line of sight images of the flame are taken in order to understand the global properties of the flame. Some representative flame luminosity images with Re jet =25,000 and 50,000 are shown in Figure 5.5. They are organized by increasing carbon number from left to right. 52 (a) Global Flame Shape for Re jet =25,000 (b) Global Flame Shape for Re jet =50,000 Figure 5.5: Impact of Re jet on flame shape with constantT coflow =1500K andS L =11.5 cm/s A couple observations are immediately apparent. First, methane has the lowest maximum luminosity and longest visible flame height. Flame height and maximum luminsoity also increase with increasing Re jet . Ethylene has a thinner and shorter flame shape in all configurations and propane,n-heptane and toluene are all very close in flame shape. The flame is always brightest at the bottom due to the presence of the pilot flame. The thickness at the tip of the flame brush also increases with increasing Re jet . Profiles of flame luminosity were taken at the centerline of the jet and these centerline profiles were used to determineH fl . The profiles are normalized by the maximum luminosity which occurs at the pilot for each flame. The normalized profiles are shown in Figure 5.6. As noticed from the flame shape images, methane has the longestH fl consistently.H fl is defined 53 Figure 5.6: Normalized flame intensity centerline profiles for various fuels at constantS L =11.5 cm/s at varying Re jet : as the location where flame luminosity drops to 25% of the maximum value away from pilot effects which is usually the secondary peak of luminsoity occuring around an axial distance of 100-200 mm. The differences in flame height are magnified at higher Re jet . There is a noticeable dip in chemiluminescence to a value near 0 which occurs at the higher Re jet ; that indicates a strong area of extinction which is in between the pilot and coflow stabilization areas. H fl will be used from this point in the global flame analysis. 5.4.2 Adiabatic Flame Height Study The first thorough characterization of the USC PPJB [40] focused on PIV and chemiluminescence- centered diagnostics of various types of fuels ofRe jet =12,500 and 25,000. There were three states that were observed: tip-ignited, tip-flickering, and tip-quenched as seen in Figure 5.7. For Re jet =25,000, the flame was tip-quenched for lower flames. Flames were only able to stabilize at near-stoichiometric or stoichiometric conditions. Therefore, higher Ka and low Da flames could not be investigated. This transition from tip-quenched to tip-ignited was sharper for higherRe jet flames. The transition from tip-quenched to tip-ignited was seen in the PIV data since the survival of the jet centerline core velocity as a function of axial distance increased with increasing . In addition to the heat provided by the central jet, the thermal expansion of the jet also acts to protect the jet centerline from mixing with the ambient air surrounding. Velocity statistics were only marginally affected by fuel type however. The major fuel effect was in the 54 Figure 5.7: Ignition Behavior of PPJB showing tip-quenched, tip-flickering, and tip-ignited states [40] globally definedH f which has been established as a marker of reactivity. This study found that when scaling the flame heightH f byS L , all of theH f ’s for the hydrocarbon fuels collapse except for methane. This is an intriguing result, especially since S L is used as an approximation when assuming flamelet-like combustion in turbulent flames. This scaling withS L as seen in Figure 5.8 informs the decision to compare between fuels with the sameS L . With this previous study in mind, the goal of the current investigation was to see whether the differences between fuels are amplified at higher Re jet and lower S L where eddies are more likely to penetrate and broaden the preheat zone for heavy hydrocarbon fuels. In order to stabilize leaner flames at higher Reynolds numbers in the tip-ignited configuration, the coflow was ignited to prevent quenching of the flame tip from heat loss and entrainment of ambient air.H fl was parametrically investigated at an unburned jet exit temperature of 298 K. This study was able to expand the investigated conditions toRe jet =25,000, 37,500, 50,000 and 75,000. The goal of this investigation will be to thoroughly characterize the global flame reactivity of the single parameterH fl as a function ofRe jet , fuel type, reactivity of the central jet (=phi), and heat loss to the surrounding coflow (T ad,jet T coflow ). First, flame heightH fl was measure for various fuels of different chemical clasisification: n- alkanes: CH 4 , C 3 H 8 andn-C 7 H 16 , branchedn-alkanes (i-C 8 H 18 , alkenes (C 6 H 12 ), cyclo-alkanes (CH 3 C 6 H 11 ), and aromatics (toluene). Both gaseous and liquid fuels were investigated withT coflow = 55 Figure 5.8: Measured scaled flame height, H f /D jet , as a function of S L for Re jet = 12,500 and Re jet = 25,000 (left and right panels, respectively). The profiles are for different fuels (see legend).) T ad,jet . Flame height is a good metric of global reactivity and therefore the flame heights were in- structive understanding the global burning characteristics of the jet flame. As mentioned earlier, the limitation of parameter space for this configuration is constrained to the empirically-tested safe limits of operation of the coflow exhaust and cooling ring of the burner which allows the coflow to reach a maximum stable operating temperature of 1850 K before harmonic resonance in the coflow begin to develop. Due to this limitation, the range of’s investigated were from =0.45-0.7 for a variety of jet Reynolds numbers. The highlights of the investigation are shown in figure 5.9 H fl ’s atRe jet =25,000 and 75,000 were plotted versus calculatedT ad,jet andS L . As seen from the piloted jet flame experiments without heated coflow, all values ofH fl between fuels collapse when plotted againstS L forRe jet =25,000. Even methane which was seen before to have a unique behvaior scales well with the other fuels of all chemical classifications when plotted versus S L . Plotting versusT ad,jet also brings the data closer together but doesn’t do as good of a job for lower T ad,jet . Now instead of being quenched at the tip,H fl decreases with increasing flame speed which is indicative of a higher reactivity of the central jet. Notably, afterT ad,jet of 1600K and aS L of 9 cm/s,H fl flattens out and stays relatively constant with increasingS L . This indicates that there is also a hydrodynamic stabilization effect on the flame height. AtRe jet =75,000, the same trends hold, however, at lower S L , there is a larger spread between data points in some cases having a difference of up to 50 mm or around 9D jet in flame height. Methane seems to have on the average a longer flame height for these measurements while toluene has the shortest. At higherS L there seems to be a difference of about 1D jet in flame height even whenH fl reaches a constant value. Since having a highT coflow suppresses the fuel-dependent differences, the effect of heat release on global flame shape was investigated next. 56 Figure 5.9: Adiabatic flame heights ofRe jet =25,000 and 75,000 scaled by calculatedT ad,jet and S L using the JetSurf mechanism 5.4.3 ExperimentalT coflow effects onH fl The effects of heat loss/heat gain are investigated by varyingT coflow . Three temperatures consid- ered:T coflow =1400K, 1500K, and 1600K. It is important to note that sinceT ad,jet is different for each fuel, the T ad,jet T coflow difference is not held constant. When T coflow is lower than the adiabatic temperature of the main jet, T ad,jet , the flame suffers heat losses (case 1); when T coflow is higher, the flame is super-adiabatically heated (case 2). S T for the first (resp. second) case is expected to be lower (resp. larger), resulting in a taller flame (resp. shorter flame). Figure 5.10a shows H fl plotted against the heat loss experienced by the jet,T ad,jet T coflow . The expected behavior is confirmed in th plot showing as H fl increases, T ad,jet T coflow also increases. The flame heights of the heavier hydrocarbon/air flames (propane,n-heptane, and toluene) align nicely when plotted againstT ad,jet T coflow . However, fuel-dependent differences are observed. Most notably, methane flames are consistently longer while ethylene flames are consistently shorter. 57 −100 0 100 200 300 30 35 40 45 50 55 T ad,jet −T coflow (H fl /D) exp Methane Ethylene Propane n-Heptane Toluene (a) Heat Loss effects 25,000 37,500 50,000 32 34 36 38 40 42 Re jet (H fl /D) exp Methane Ethylene Propane n-Heptane Toluene (b) Re jet effects 30 35 40 45 50 55 30 35 40 45 50 55 (H fl /D) exp (H fl /D) LES Methane Ethylene n-Heptane Toluene Ideal Behavior (c) Simulations vs. experiments Figure 5.10: Impact of T coflow at a fixed Re jet (left) and Re jet at a fixed T coflow = T ad,jet (center) on the experimentally-measured flame heights. To emphasize the almost linear dependence on Re jet , a linear fit is shown in dashed lines. Comparison ofH fl predicted by the LES and measured experimentally at the three different Re jet ’s forT coflow =1500K (right). 5.4.4 Experimental Re jet effects onH fl Next, the effects of Re jet are shown in Fig. 5.10b). Note, all experiments are performed at adiabatic conditions (T coflow = T ad,jet ) in order to remove the effects of heat losses. WhenU jet (equivalently Re jet ) is increased at a fixed S L , the flame is expected to be longer as seen in the original PPJB experiments [28]. Assuming a constant turbulent flame speed,S T , along the time-averaged flame front, the following scaling is expected: H fl D U 2S T : (5.4.1) 58 This behavior is observed consistently for all fuels, as H fl increases fairly linearly with Re jet . Methane has the largestH fl , ethylene has the smallestH fl , and the rest of the fuels fall somewhere in between. More interestingly, all fuels but methane exhibit the same slope (within experimental uncertainties). One may conclude that, under the present conditions,S T ’s should be very similar for all the fuels investigated with the exception of methane. 5.4.5 Experimental and numerical comparison ofH fl Differences in H fl were magnified with increasing Re jet , simulations and experiments are com- pared with one of the highest Re jet cases (but not so high that H fl can no longer be defined). Without surprises, hydrodynamic effects represented by Re jet are well captured in the LES. As shown in Fig. 5.10c), predictedH fl from the LES are within 10-15% of the experimental values. The experimental trends of longer flames for methane and toluene thann-heptane and ethylene are reproduced. These results are especially interesting given the assumption behind the combustion models, namely that the structure of the turbulent flame is the same as that of a laminar flame. With this modeling restriction, the systematic, albeit small, differences between predictions and measurements may be viewed as indirect signs that further modeling development are needed. 5.5 Conclusions This study investigated fuel and hydrodynamic effects on piloted premixed jet flames. Particularly, the assumption that a constant laminar flame speed is sufficient to model the global flame structure was tested. This investigation was carried out at a constant laminar flame speed to isolate these effects. The parameters varied include heat losses, jet Reynolds number, fuel molecular weight, and fuel chemical classification. Flame heights defined based on CH* chemiluminescence were used as metric to compare the different flames. From this extensive parametric study, H fl was shown to be a good marker of reactivity. Fuel effects are supressed under adiabatic coflow conditions and can be approximated with the laminar flame speedS L . This holds true even at the highest jet exit velocity of 200 m/s and Re jet =75,000.H fl showed the widest differences between fuels at low laminar flame speeds below the flammability limit so laminar flame speed is not a good marker at the limits of flammability. Experimental results indicated that small amounts of heat loss may play a significant role on the jet reactivity as the flame heights scaled with the heat loss from the jet. However, differences between flames with different fuels could still be seen in the absence of heat loss, and these differ- ences were magnified at higher Reynolds numbers. Particularly, at the same laminar flame speed, methane flames were consistently taller and ethylene flames were consistently shorter while other fuels had approximately the same height. LES reproduced the experimentally observed trends in 59 global flame heights (effects of heat losses and Reynolds number) but some of the differences between fuels were not captured. Boundary conditions of the velocity flowfield were measured using PIV for the experimental configuration. Matching LES were performed of the same experiments and were validated against 2D PIV measurements. The characterization of the boundary conditions of temperature and veloc- ity of the jet are important in order to provide a reasonable starting point for modelers to match the flow properties. These measurements provide a database of flame properties for the turbulent jet flame configuration. 60 Chapter 6 Experimental Estimation of Turbulent Flame Speed and Characterization of Global and Local Observables for Lean Jet Fuel/Air Mixtures at High Ka 6.1 Introduction The turbulent flame speed,S T , has been widely studied as it correlates directly to the burning and heat release rates in combustion devices. Early work by Abdel-Gayed and Bradley [88] examined a wide array of experimental configurations and identified that the ratio ofS T to the laminar burning velocity, S L , could be correlated with the ratio of the r.m.s velocity, u 0 , and S L . A dependence on the cold gas turbulent Reynolds number was also noted, as S T was shown to increase with higher values ofu 0 . However, after a near-linear increase withS L increases inu 0 caused a slowing of the rate of increase ofS T , referred to as the bending phenomenon. Similar observations were addressed by Filatyev [89], Duclos [90], and Peters [9], among others. In their comprehensive literature reviews, Lipatnikov and Chomiak [14, 91] discussed turbulent flame speed evaluation methods, highlighting the use of turbulence statistics for scaling parameters. It is desirable to correlateS T with a relatively small number of turbulent flow and combustion parameters. In view of this consideration, all experiments aim to derive a singleS T values with the underlying assumption being thatS T along the flame brush is constant. However, experiments and computations have shown that upstream wrinkling is coupled with the downstream wrinkling of the flame [13]. Flame wrinkling and area generation are often correlated to increases inS T . Thus calculations of S T are highly dependent on the experimental configuration. Driscoll argues that while homogenous isotropic turbulence appears desirable, it has not been demonstrated that such conditions could be created nor that it would be of relevance to real environments. To account for this geometric dependence, turbulent flame studies are categorized into four types of experimental configurations: envelope flames, oblique flames, flat flames, and spherical flames. S T is usually defined as a global consumption speed, which is a measure of overall mass con- sumption, local consumption speed, or a displacement speed, which is a measure of the propaga- tion speed of the flame brush. Driscoll [13] proposed that each definition provides a valid method 61 to compare experimental results against a simulation using identical geometry. The global con- sumption speedS T,GC provides a relationship between the mass flow of reactants to the flame and the average flame surface area, most commonly defined at the location of mean progress variable contour c = 0.5 [13]. S T,GC = _ m r r A c=0:5 (6.1.1) Second, the global consumption speed is commonly used in the Bunsen flame configuration, as all reactants must past through the flame brush and therefore are consumed by the flame. This requirement is not met by counterflow or V-flames. Local consumption speedS T,LC provides a local parameter by relating the laminar flame speed, a turbulent stretch factor, and the flame surface density , as discussed by Bray and Cant [65]. S T,LC =S L0 I 0 Z 1 1 d (6.1.2) is defined as the averaged area of flame surface per unit volume and must be experimentally determined using two-dimensional approximations. Shepherd [92] suggests that can be esti- mated by relating the area between two c contoursA with the averaged length of the flame front contours L within that region: = L=A. Measurements for have been completed in a wide variety of experimental configurations. Local displacement speedS T,LD , which relates the flame velocity V flame to the incoming reactant velocity V gas normal to the leading edge. S T,LD = (V flame V gas ) n LE (6.1.3) Local displacement speeds are commonly defined in counterflow, spherical and V-flames, as V flame goes to zero for stationary flames. Cho and Law [93] utilized laser Doppler velocimetry (LDV) in a turbulent stagnation flow configuration to provide a comparative local turbulent flame speed. Both local consumption and displacement speed measurements require the determination of the flame leading edge contour. However, this contour has not been consistently defined among experimental studies, complicating efforts to compare between different studies and configurations. Moreover, the three turbulent flame speed quantities are not equivalent and should not be compared against each other, as shown by Shepherd and Cheng [94]. Though turbulent flame speeds have been defined using a variety of experimental configura- tions, they have been focused on relatively low Reynolds number and Karlovitz number conditions. However, under high Reynolds and Karlovitz number conditions, a different regime of turbulent combustion is accessed, as shown in the turbulent regime diagram proposed by Borghi [10] and Peters [8]. In this regime, the smallest flow eddies can enter the preheat zone of the flame and even 62 cause the reaction zone to be broken [29]. However, a consensus has not been reached to explain how increasing turbulence levels affects the thickness of the turbulent flame front, and new ex- perimental apparatuses may facilitate further exploration of conditions above the Klimov-Wiliams criterion [95]. The development of the piloted premixed jet burner (PPJB) [96] allows access to the thin and broken reaction zone regime as characterized by high turbulent Reynolds and Karlovitz numbers. The authors have developed a modified PPJB facility to reach intensely turbulent conditions. Given the steady-state nature of the jet flame, this study will focus on additional means to determine S T both global and local by utilizing detailed measurements of the flow velocities. Additional flowfield statistics of the full jet flame and resolution of the full energy spectra will be attempted in order to provide more detailed turbulence properties at the location of the flame surface. The focus of this study was to fully characterize the velocity field statistics of the burner and provide a database of S T data as well as investigate additional global flame observables such as flame height which has been established in previous studies. While much of the previous effort has been to characterize how hydrocarbon flames of different molecular weights and chemical structures interact with highly-turbulent flows, the current study extends this investigation to Jet A which is used in high speed airbreathing propulsion devices. 6.2 Experimental Approach 6.2.1 Burner Operating Conditions The Piloted Premixed Jet Burner apparatus developed at the University of Southern California [40, 87] was used in this study. A constant =0.80 ethylene/air pilot with an exit velocity of 0.75 m/s was kept as the constant flame stabilization mechanism in all cases. The pilot was changed to an ethylene flame to decrease the maximum temperature at the base of the burner and to provide a three-stream flow with lean conditions in each stream. As mentioned in previous studies, a mixture of H 2 /C 2 H 4 /air in the coflow was found to have the best stability and flexibility of operation and enabled the burner to stay running for extended periods of time. There were two types of investi- gations performed in this experiment. For the parametric flame heightH fl study, the coflow was run atT coflow =T ad,jet . Additional studies were performed for higher at a fixed value ofT coflow for measurements of flame height, temperature and global flame shape at a fixedS L . In order to premix and pre-vaporize lean jet fuels and prevent fuel condensation, the fuel/air mixture in the vaporiza- tion system was heated to 403 K. More details on the vaporization system are described in Chapter 2. The jet Reynolds numberRe jet was kept the same as previous studies at anRe jet =25,000 and 50,000. Due to the higher u , however, the exit velocity was increased toU jet =112 m/s and 225 m/s. As in previous studies, methane, ethylene,n-heptane, and toluene were investigated. Jet A, 63 JP-8 and JP-5 were also investigated. S L andT ad;jet of these mixtures used to scale the results were calculated using a novel lumped mechanism for jet fuels which is still able to capture the effects of decomposition called HyChem and using the PREMIX and EQUIL codes. [4] As in previous studies, these experiments fall between the thin to broken reaction zone regime of the turbulent premixed regime diagram. 6.2.2 Chemiluminescence Flame Shape and Flame Height As discussed in previous studies [40, 87], mean chemiluminescence images can be used as a marker of global flame reactivity. Flame heights (H fl ) based off of the mean centerline intensity were measured for a variety of flames including the three types of jet fuels: Jet A, JP-8, and JP-5. 50 images of each flame were taken with a 16 bit Andor Zyla camera coupled to a 50 mm Nikon lens with a resolution of 0.043 mm/pixel and an exposure time of 0.1 ms and images were averaged together. These images are then Abel-inverted as in the work by Sidey and Mastorakos [97]. Global Consumption Speed from CH* Global consumption speed based on the mean chemiluminescence-based flame brush, S T,GC,CH* can be calculated using jet inlet areaA jet , jet velocityU jet , and jet surface areaA jet,flame as shown in the following equation: S T,GC,CH* =V jet A jet,inlet A jet,CH* : (6.2.1) Figure 6.1: Binarized flame brush vs. extracted flame perimeter used to calculatedS T,eff A jet,flame was estimated by first extracting the edge of the mean chemiluminescence flame shape in MATLAB by binarizing the chemiluminescence image and then extracting an edge from the 64 MATLAB perimeter function as shown in Fig. 6.1. The flame surface area was computed assuming an axisymmetric jet. Note: this method is not designed for exact analysis but comparative estimates of flame speed. For more precies estimates, the instantaneous flame surface area would have to be visualized as in the Mie Scattering method mentioned in a following section. 6.2.3 Temperature Measurements Temperature measurements along the centerline are taken with a ceramically-shielded thermocou- ple (Omega, Inc) and corrected for radiation. The thermocouple tip, 200m in diameter, is inserted into the flame perpendicular to the incoming flow with approximately 1 mm of the leads exposed. The thermocouple measurements have an uncertainty of50K everywhere except at the interface between flows. There, the uncertainties reach100K due to oscillations in the measured temper- ature values. 6.3 Experimental/numerical validation 6.3.1 Laser-Based Diagnostics An extensive Particle Image Velocimetry (PIV) analysis of the flames was performed using a vari- ety of camera and laser configurations and tracer particles as detailed in the sections below. Both PIV and Mie scattering imaging techniques are utilized in the experiments. Stitched PIV To provide improved resolution in the imaging, a stitched camera configuration is used. The method shown in Figure 6.2. Three 16-bit Andor Zyla cameras with a resolution of 2160 x 1400 pixels are positioned with a preset vertical overlap to capture the length of the jet, and the subsequent images are ”stitched” together using MATLAB. The three-camera stitched configuration provides a resolution roughly three-times the resolution one camera for the same domain of interest. Though this stitched con- figuration currently uses three cameras, additional cameras could be used to provide even greater resolution. This configuration can capture from the exit of the burner up to 350 mm above the burner exit. The stitched PIV images were taken for selected fuels at the conditions listed in Table 6.1 in order to determine the turbulence properties of the burner. PIV images are captured using a laserline filter centered on 532 nm and focused using a F85 mm/2.8 Nikon lens for a resolution of 56m/pixel. Green light at 532 nm is generated at 10 Hz using a frequency doubled, dual cavity Q-switched Nd:YAG laser (Brilliant Twins, Quantel). The 65 Figure 6.2: Stitched PIV Method beam is expanded using a plano-cylindrical lens to illuminate the tracer particles in a plane crossing the burner center axis. Aluminum oxide particles of 0.3m diameter are used to provide full flow field statistics, as the tracer particles survive the flame. Table 6.1: Measured and calculated scaling parameters for investigated PPJB conditions. Integral length andU 0 were measured at a distance ofx=D = 15 in the shear layer of the flame. 1000 statistically independent image pairs are captured for each data set, using an inter-frame time of 3-6s. The images are corrected by subtracting the average background measured without particle seeding and then processed using the Davis Flowmaster software (LaVision Research Inc.). Velocity fields are generated using 50% overlapping interrogation windows of decreasing sizes, down to 12x12 pixels, along with filtering and smoothing between passes to minimize spurious results in the multi-pass cross-correlation algorithm. This gives 0.4 mm/vector final resolution for the flowfield. Turbulent Burning Velocity and Mie Scattering As mentioned earlier, the global consumption speed is commonly used with Bunsen flames, as all reactants must past through the flame brush. The PPJB provides a similarly convenient way of measuring the global consumption speed. The instantaneous flame surfaces can be measured using the same thin laser sheet but illuminating silicon oil particles, which act as a constant temperature contour and marker for the flame front. Silicon oil particles evaporate at 473-523 K so these 0.3m 66 diameter particles can also be used as flow tracers in order to provide velocity vectors exclusively from the unburned mixture [15]. It was used in this experiment to provide a description of the instantaneous flame surface and determine mean progress variable contours. (a) Raw Particle Image (b) Binarized Surface (c) Surface Veri- fication Figure 6.3: Post-processing technique to determine instantaneous flame surface from Mie scatter- ing images 1,000 statistically independent Mie scattering image pairs were captured and then filtered to remove digital noise, using a filter size near the inner cutoff scale. The filtered images were bina- rized using a threshold intensity [98] to produce an instantaneous progress variable field, as shown in Figure 6.3(b). This binarized surface was compared to the instantaneous Mie scattering image to confirm that the chosen threshold value reproduces the raw surface, as shown in Figure 6.3(c). The instantaneous progress variable snapshots were averaged to produce a mean progress variable field c exp with intensity gradient of 1 to 0, corresponding to unburnt and burnt reactants respectively. The averaged experimental flame surface was then defined using the c exp =0.5 isocontour. [92] The instantaneous isosurfaces are then averaged to produce a c exp field. The averaged computational flame surface is calculated using the c exp =0.5 isocontour. The c exp =0.5 isosurface will be used as an average flame surface calculated in a similar manner toS T,GC,CH* for the turbulent flame speed calculations. The equation for global turbulent consumption speed calculated from Mie Scattering is shown below: S T,GC; c=0:5 =V jet A jet,inlet A jet; c=0:5 : (6.3.1) 67 Nested PIV Although the stitched PIV provides better resolution than the single camera setup, the turbulent jet flame encompasses a wide range of scales: from large scales on the order of the jet diameter around 6 mm to the smallest Kolmogorov scales of 7m. Due to the large differences in the range of scales, it is impossible to simultaneously resolve the entire jet flow field as well as the energy spectra with the stitched PIV setup. However, since energy is dissipated at the Kolmogorov scale, the only way to resolve the dissipation spectra as well as the generation spectra is to zoom in to an extremely fine resolution while simultaneously taking data of the full jet. This novel technique called Nested PIV has only been attempted in one other configuration to the author’s knowledge [99]. However, the Nested PIV technique is being attempted for the first time in the presence of reacting flow. Figure 6.4: Nested PIV Example Image The 3-camera nested PIV setup focusedx=D of around 12 in order to ensure capturing defini- tively burned and unburned statistics and testing the limits of the hierarchical setup. The setup focused on three different fuels with a variety of to try to understand the effect of varying jet reactivity and fuel type on the velocity fields of the flame: Jet A, ethylene, and methane. Fig- ure 6.4 shows an example of three instantaneous PIV images taken in this setup. The setup had a resolution of 42.4m/pixel down to 3.8m/pixel for the most zoomed-in window as described in Table 6.2. The setup used a ND-YLF laser centered at 527 nm and put through beam expanding 68 optics to create a thin laser sheet at the jet centerline to illuminate 0.3 micron alumina tracer parti- cles. The laser was operated at a frequency of 25 Hz for statistically independent images. As it was impossible to align three cameras along the same axis with the desired field of view, a scheimpflug configuration was used for the zoomed out camera and all images were corrected for distortion with a calibration target and a distortion map generated by the calibration in Davis Flowmaster. A 200 mm zoom Nikon lens was used with a lens tube in order to zoom into the smallest resolution for the flame. A 50 mm fixed focal length Nikon lens with an aperture of 4D was used for the full frame image. The middle camera used a 105 mm fixed focal length lens. Images were post- processed with the Davis Flowmaster software with final interrogation window sizes of 16x16 and 8x8 pixels. Universal outlier detection was used to remove spurious vectors between passes. 1000 statistically independent images were averaged to obtain 2D flowfield statistics. Table 6.2: Resolution and Window Size of Nested PIV camera setup With a single laser, there is a tradeoff to be made when using this method. The ideal pixel displacement has been found to be typically around 6-10 pixels for good correlation with the PIV post-processing algorithm from previous testing. However, due to the differences in resolution between the images, it is impossible to have one dt that is optimal for all cameras. Therefore, the full frame PIV skews toward a lower pixel displacement while the zoomed-in PIV images have a larger pixel displacement. A dt optimization was done for each configuration to ensure the average and r.m.s. velocity fields were similar for more optimal dt values. It was found that even with a maximum pixel displacement of 2 pixels, the correlation algorithm was still able to reproduce the mean velocity field. As a compromise between all of the ideal pixel displacements, a dt of 0.75s was chosen for aU jet =150 m/s. However, this corresponded to a pixel displacement of around 2 pixels for the mean flow in the largest FOV causing a mean error in velocity of around 20% due to the limitation in pixel resolution. This also corresponded to a maximum 15 pixel displacement for the most zoomed-in window. While this percentage is relatively high, the overall flowfield can still be reproduced since PIV can capture sub-pixel displacements based on texture. 69 Particle Image Density for Conditional Statistics Although liquid particles can measure the unburned properties of the flame, it is instructive to understand the behavior before and after the flame. To this end, an edge detection technique using solid alumina particles was attempted in order to measure burned vs. unburned statistics. Higher concentration of particles in the unburned region of the flame results in a higher image intensity in the center of the jet. Subsequent expansion and mixing of the gas with the surrounding coflow in the burned region of the flame causes a lower gas density, which is correlated to lower mean particle image intensity. Figure 6.5: Image showing burned vs unburned binary mask created from a particle density image This program, which is written in MATLAB, first pre-processes the raw PIV images through a median and Gaussian filter to remove the peaks from scattered light off of the larger particles. All information above 2.5 standard deviations of the mean particle intensity is automatically assumed to be in the unburned region of the flame. Then, the image was scaled from 0 to 1, 1 being the strongest intensity representing an unburned particle density, and 0 representing an absence of particle intensity. The image was then binarized based on different particle image intensity levels. Since there is a sharp cutoff in intensity at the flame surface, the cutoff for the unburned statistics was defined at 0.5 of the maximum intensity. The cutoff for particle intensity of the burned jet statistics was between 0.10.5 of the maxi- mum intensity for each image. These images were then binarized and applied as masks to the PIV images. These masked velocity fields were then averaged to determine conditional statistics at the burned vs. unburned surfaces. An example of the typical particle image mask created from this algorithm is shown in Figure 6.5. The burned vs. unburned statistics can only be created for the largest viewing window because when zooming into the flame, it is harder to determine a concrete edge due to resolution of more vectors and subtle particle density variation. Creating a particle 70 density map for each image may be a better alternative for future studies. 6.4 Global Flame Observables: Results and Discussion 6.4.1 Fuel Effects on Global Flame Shape Non-adiabatic flame shapes at matchingS L = 15.3 cm/s were conducted for the conditions listed in Table 6.3. The jet fuels mentioned in the table include Jet A, JP-8, and JP-5 and were investigated with the addition of heat loss to magnify differences between the flames so the coflow was kept at a constant temperature ofT coflow =1500 K. Table 6.3: Condition matrix for flame heights of jet fuels The global flame shapes of integrated line-of-sight CH* luminosity were examined for Re jet =25,000 and 50,000 in Figure 6.6. It is apparent from the images that ethylene has a noticeably different flame shape than the rest of the flames. For Re jet =25,000, the ethylene flame is much shorter and the flame brush is thinner indicating that it may be in a different regime of combustion. For Re jet =50,000, ethylene starts to develop the more characteristic thicker brush shape. All flames exhibit an area of strong extinction right after an intensity maximum due to stabilization from the pilot. Interestingly, all of the jet fuels have approximately the same flame shape and length. JP-5 has the highest peak chemiluminescence due to the presence of higher percentages of aromatics. n-Dodecane, which is typically used as a one-component surrogate for jet fuel, has a slightly shorter flame height than the jet fuels though a fairly similar flame shape. Methane has the lowest chemiluminescence consistently among the different fuels explored. With these images in mind, the profiles of chemiluminescence at the centerline scaled against maximum intensity of methane are plotted in Figure 6.7. The maximum intensity occurs close to the jet exit due to the interaction of the jet flame with the pilot flame. This maximum value is dependent on the central jet fuel. It is interesting to note that JP-5 has a lower maximum inten- sity than toluene at Re jet =25,000 but has a higher maximum value for Re jet =50,000. With an increase in Re jet , Jet A and JP-8 start to have almost the same exact intensity distribution along the centerline as do then-alkanesn-heptane andn-dodecane. Compared to the intensity of methane, 71 (a) Global Flame Shape for Re jet =25,000 (b) Global Flame Shape for Re jet =50,000 Figure 6.6: Impact of Re jet on flame shape of Practical Fuels with Constant T coflow =1500K and S L =15.3 cm/s the maximum intensity of ethylene and toluene stay fairly constant and may even decrease slightly. However, there is a relative increase in maximum CH* production for then-alkanes and jet fuels which could indicate an overall increase in reactivity due to the presence of turbulence. Since these flame brush shapes and centerline profiles are in the presence of varying amounts of heat loss, adiabatic flame height studies were conducted to see if these trends could be reproduced. 72 Figure 6.7: Centerline scaled intensity profiles of jet fuels 6.4.2 Adiabatic Flame Height Study Flame heights, H fl , forRe jet =25,000 and 50,000 were profiled from mean chemiluminescence images of various fuels of = 0.40.7 with ignited coflow at T coflow = T ad,jet . The results are plotted against laminar flame speedS L as in previous studies. As a reminder,H f l has been found to scale well with S L . S L for Jet A is calculated with the HyChem mechanism [4]. Results are shown in Figure 6.8. Figure 6.8: Flame heights (left) and maximum chemiluminescence intensities (right) of lean pre- mixed jet flames adiabatic coflow and varying fuel type. The top two plots are atRe jet =25,000 and the bottom plots are atRe jet =50,000 73 As found from previous studies, H fl decreases with increasingS L . However, it does not de- crease linearly, but starts to flatten out at a constant value around an S L of 15 cm/s. With the addition of ethylene, there is a marked difference between the stabilization height of ethylene ver- sus the other fuels as seen in Figure 6.8. At lowerS L ,H fl increases rapidly and at least for methane and toluene, the values seem to diverge. Jet A follows exactly the same behavior as then-alkanes propane andn-heptane. Looking at the maximum camera intensity of CH,I max on the right, in- tensity of CH* increases fairly linearly for all fuels. Interestingly,I max for Jet A falls in between n-heptane and toluene. Figure 6.9: Maximum CH* Mass Fraction for Various Fuels from 1D Laminar Flame Calculations 1D laminar flame calculations duplicate these results. It can be seen from the path analysis of these flames, that the CH* produced in n-alkanes is mainly produced from the breakdown of the fuel into n-alkenes which finally break into C 2 H 2 and C 3 H 3 which produce C 2 H and finally CH*. Toluene has a different path of production starting from the breakdown of the benzene ring which breaks down quickly into C 3 H 3 which breaks down to C 2 H which produces CH*. This leads to an overall higher value of CH* produced by aromatics as seen by the high CH* intensity for toluene. Although Jet A is mainly composed of n-parrafins, the presence of a small amount of aromatic species causes it to have a factor of 2 increase in intensity over the n-alkanes. The laminar calculations of maximum CH* are shown in Figure 6.9 for reference. The jet flame configuration is an envelope flame as discussed earlier. This means that a global consumption speed can be defined based on the mean flame area assuming all of the mass flowrate of the inlet passes through the flame surface. A flame surface is typically defined from averaging binarized images of the reaction surface from Mie scattering (which will be discussed in the fol- lowing section). However as CH* typically falls in the reaction zone of the flame, the flame surface is defined from the integrated CH* surface as discussed in the experimental methods section. The results of this investigation ofS T,GC,CH* is show in in Figure 6.10. On the left is a plot ofS T,GC =S L showing a linear scaling withU 0 =S L . On the right its the value ofS T,GC,CH* vs.H fl . As expected,S T,GC,CH* is proportional to the inverse ofH fl . With a shorterH fl 74 Figure 6.10:S T,eff for multiple fuels and correlation withH fl comes a smaller average surface area and therefore a higher flame speed. This indicates also that H fl scales fairly linearly with S T,GC,CH* so the information captured by H fl describes the global flame reactivity from the perspective of CH* reaction contours. 6.4.3 Mie Scattering Results and Discussion The global consumption speed measurement from Mie scattering is explored to provide a global perspective on the jet flame behavior. S T,GC; c=0:5 is evaluated using the c field as discussed in Section 6.3.1. The experimental c exp =0.5 isosurface is then binarized. The average flame surface area is then calculated from this binary surface, assuming axisymmetry in a similar method to the area calculated forS T,GC,CH* as discussed in Section 6.2.2. (a) Binarized Flame Area 0 5 10 15 20 25 30 35 40 0 200 400 600 800 1,000 0 5 10 15 20 25 30 35 40 0 0.2 0.4 0.6 0.8 1 1.2 x/D CH* Intensity (counts) Mean Progress Variable (¯ c) CH* ¯ c (b) Centerline Profiles Figure 6.11: Binarized Average Flame Surfaces produced from Mie Scattering (dashed line) and CH* (white area) (left) and comparison of CH* intensity vs. c at the centerline (right) S T,GC; c=0:5 was calculated and compared toS T,GC,CH* for selected cases to try to understand the 75 relevance of the two different parameters. An example of a comparison between the two surfaces generated from CH* vs Mie scattering and the difference at the centerline is shown in Figure 6.11. The figure reveals a very different flame shape between the Mie Scattering surface vs. CH*. First of all, the CH* height is about a factor of two larger and the surface has an expanding then contracting shape. The c =0.5 surface is more conical in shape and is much shorter and steadily decreases in radius as function of axial distance until the flame tip resembling a typical bunsen flame shape. These two surfaces can be representative of the inner and outer bounds of reaction of the flame. S T,GC; c=0:5 andS T,GC,CH* were calculated for ethylene, methane, and Jet A at Re jet =25,000 and the sameS L =30.3 cm/s and adiabatic conditions for the coflow. (a) Silicon vs. CHH fl (b) Instantaneous Flame Perimeter P (c) S T,GC,CH* (d) S T,GC; c=0:5 Figure 6.12:S T,GC; c=0:5 andS T,GC,CH* for Methane, Ethylene, and Jet A A comparison ofS T,GC; c=0:5 andS T,GC,CH* is shown in Figure 6.12. Figure 6.12a) first shows the flame heights obtained from the binarized images. As seen from the image above,H fl from CH* is more than double the value ofH fl from the Mie scattering surface. However, the fuel-dependent differences are preserved in that C 2 H 4 has the fastest turbulent flame speed while methane is the least reactive with the lowestS T,GC for both parameters. Figure 6.12c) and d) compare the scaled S T,GC =S L for the two methods.S T,GC,CH* has a much smaller magnitude since it is only around 3S L . Also the % differences between ethylene and methaneS T,GC,CH* is only around 13% in comparison to around 20% forS T,GC; c=0:5 even though the differences inH fl were larger. This is due to the fact that the flame area is much larger for the CH* surface. The Mie scattering surfaces have a values of around 7.5-9S L . 76 Another parameter that can be calculated from the Mie scattering contours is the averaged in- stantaneous flame perimeter which is shown in Fig. 6.12b). This helps to account for wrinkling on the flame surface which is not considered in the average flame brush. The instantaneous perimeter is about 2x longer on average than the averaged perimeter. It follows the same trend asS T,GC which means that C 2 H 4 has the most wrinkled instantaneous surface while CH 4 is the least wrinkled. This agrees with a commonly made assumption that an increase in turbulent burning velocity typically corresponds to an increase in wrinkled flame area as hypothesized by Damkohler [12]. A comparison of the two methods for calculatingS T,GC shows that the simple CH* luminosity measurement can capture the overall trends in reactivity throughS T,GC andH fl . However, these fuel differences are more subtle when using CH* as compared to Mie Scattering. A case can be made that a lot that much can be learned from simple chemiluminescence imaging. 6.5 Temperature Measurements Temperature measurements were performed along the burner centerline and selected radial profiles were taken as well in order to determine the averaged temperature density for different flames. Temperature measurements were taken for unburned jet at 403 K with ignited pilot and coflow, and CH 4 /air and C 2 H 4 /air along the burner for a fixedS L =15.3 cm/s corresponding to =0.54 for the methane/air flame and =0.43 for ethylene. These conditions corresponded to the stitched PIV conditions that will be mentioned in a following section. A previous study has already determined that the temperatures at the jet exit are almost equivalent for all mixtures since there is very little entrainment and mixing at the jet exit. This study will focus on the temperature profiles at higher axial distances. Axial profiles of temperature are plotted in Figure 6.13. (a) Axial Temperature Profiles of CH 4 /air, C 2 H 4 /air, and air jet with 1500 K coflow (b) Axial Temperature Profiles of CH 4 /air jet flame at 1500K and 1600K Figure 6.13: Axial Temperature Profiles Figure 6.13a) shows temperature profiles of CH 4 /air, C 2 H 4 /air, and air jet with 1500 K coflow. 77 The cold air case is the baseline temperature field into which the flame propagates. For the cold jet, the centerline temperature starts to increase at aboutx=D =7 and then increases due to mix- ing and radiative heat transfer from the surrounding coflow. Now, comparing the cold jet case to the reacting case, the temperature profiles match up to an axial distance ofx=D =12 and then the temperature begins to diverge and the ignited flames increase linearly to the maximum temperature value. Since methane has a higherT ad,jet than ethylene, the maximum value is higher for methane. The maximum value then decreases down to 1500 K, the temperature of the coflow. The temper- ature increase is gradual is occurs for a total distance of around 15D. The peak for both flames is aroundx=D around 25-30 althoughH fl is shorter for ethylene. Figure 6.14: Radial Temperature Profiles Figure 6.13b) shows a comparison between temperature profiles of methane/air flames atT coflow =1666 K and 1500 K. The peak of the temperature profile occurs at the same place for both flames. T coflow =1666 K a higher temperature atx=D =10 however, the overall shape is fairly similar af- terx=D =25 except for around a 150 K temperature difference corresponding to the difference in coflow temperature. Radial profiles of temperature were also measured atx=D =10 and 20 as seen in Figure 6.14. It is also interesting to note that the 523 K average value which is the temperature contour where silicon oil evaporates seems to be around x=D =10 for all flames as well as the pilot. This indicates that Mie scatteringH fl should be shorter than theH fl from the CH* reaction layer which occurs atx=D =20-30. Radial temperature profiles atx=D =10 show that the peak of ethylene and methane are very similar in value and have a peak in temperature around r=D =1 before decreasing down to the temperature of the coflow. The air jet steadily increases from around 500 K to 1500 K in the coflow. Atx=D =20 close to the location of maximum flame intensity there are marked differences between methane and ethylene. This time even the cold jet flow is preheated to 1000 K on average in the centerline indicating that mixing and heat transfer are very prominent at this location. The peak of temperature for ethylene is slightly closer to the centerline for ethylene which agrees with experimental observations that the average CH* jet flame brush shape is thinner for ethylene than 78 for methane. Additionally, ethylene has a higher average temperature even though it has a lower at the centerline indicating that flame is flickering in and out of this location more frequently than methane. This also agrees with the CH* measurements sinceH fl is shorter for ethylene than for methane. 6.6 PIV Results and Discussion 6.6.1 Stitched PIV The goal of the stitched PIV was to demonstrate well-resolved velocity statistics for the entire jet flame and characterize the statistical properties of the burner and determine if there are visible differences in the velocity profiles for different fuels that are visible with this velocity resolution. Also, the goal was to determine local turbulence properties to understand how the flame behavior might change along the jet axis. To accomplish this, methane, ethylene, and Jet A were investi- gated at anS L =15.3 cm/s forRe jet =25,000 and 50,000. Additionally pilot and air cases were investigated in which the pilot case represents ignited pilot and coflow but only air in the jet while the air case injects air in all three streams.The centerline profiles of mean, U x and r.m.s velocity,u 0 are plotted in Figure 6.15. The centerline mean velocity profiles forRe jet =25,000 are almost identical for cold jet with ignited coflow and ignited flames. There is almost no effect from the presence of the flame. Com- pared to the cold jet, however, the mean velocity displays a much different behavior. This shows that the temperature of the surrounding coflow and pilot are the most responsible for the velocity profile shape and that the flame should be compared to the ignited coflow case instead. u 0 values of the flowfield are important since the value ofU 0 is often used as a scaling value for turbulent flame speed. u 0 values are very similar for all of the fuels with ethylene having a slightly higher peak. This was surprising since ethylene had a much lower flame height. However, since S L was close to the limits of reactivity at 15.3 cm/s and T ad,jet was very close to T coflow , it is possible that since the velocity field is largely influenced by temperature, there is very little effect of the flame on the velocity field. However, the peak of the turbulence intensity is around anx=D =20 which is close to the location of maximum intensity for CH*. The cold airflow case has a peak much closer to an axial distance of 10D. All of these differences in the velocity fields are very subtle. Differences in the velocity fields can be magnified when looking at higher order moments. Therefore, the turbulent kinetic energy (TKE) is shown for the 2D profiles to get a better idea of how the flame shape compares to kinetic energy generation in the velocity fields as seen in Figure 6.16. 79 Figure 6.15: Stitched PIV Centerline Profiles. The pilot case represents ignited pilot and coflow but only air in the jet while the air case injects air in all three streams. TKE is calculated from the following equation: TKE = 2 X i=1 3 4 (U i U i ) 2 (6.6.1) The value of 3/4 assumes that the third component of velocity is on the same order of magnitude of U r and U x . The 2D TKE maps show that the all configurations have TKE generation in the shear layer. However, the magnitude of TKE is lower for the cold jet than the flames indicating that there is TKE generation from the presence of the flame. At anx=D around 20, ethylene has a much larger peak in TKE than methane and the TKE layers are closer together. This is due to the presence of the flame since the flame height of C 2 H 4 is shorter than both methane and Jet A. Jet A has a very similar behavior in velocity to methane for these cases since it also has a very similar H fl . Another parameter of interest is the integral length scale,L 0 which is used to calculate turbulent Reynolds number, turbulent time scale and other turbulence properties. L 0 is calculated at each 80 Figure 6.16: Stitched 2D Turbulent Kinetic Energy Profiles location by using a two-point correlation of the velocity field using the following equations: L ii = Z <U 0 i (~ x)U 0 i (~ x +~ r)> <U 0 i (~ xU 0 i (~ x)> dr (6.6.2) L 0 (L xx (~ r 1 )L xx (~ r 2 ) +L rr (~ r 1 )L rr (~ r 2 )) 1=2 (6.6.3) where ~ r 1 is the horizontal direction and ~ r 2 is the vertical direction. Figure 6.17:L int along the centerline and at a radial distance of x/D =15 Figure 6.17 shows L 0 in the axial direction at the centerline and a radial profile of integral length atx=D =15 where the statistics are taken for the jet flame in order to place the flame on the premixed Borghi diagram. Integral length increases linearly along the jet axis and has a similar linear behavior for both Re jet . Profiles of L 0 at x=D =15 show that the integral length is fairly 81 constant across the jet and is about a factor of 2 lower for the higherRe jet . Because the profile of integral length is constant across the jet, one value ofL 0 can be associated with each value of axial distance. The linear increase in L 0 makes intuitive sense since the jet half width also increases linearly as a function ofx=D so as the jet spreads,L 0 also increases. Figure 6.18:Ka andRe t calculated at the burner centerline WithL 0 andu 0 measured and defined at each axial location, local non-dimensional turbulence parameters of Karlovitz number Ka and turbulent Reynolds number Re t can be calculated. As mentioned earlier, these values are instructive in understanding how the flame will interact with the local turbulent flowfield. As a reminder, Ka is defined as t f =t , the ratio of the chemical timescale to the timescale of the smallest eddies in the flow. Values ofKa>1 indicate that eddies on the order of the Kolmogorov scale of dissipation can interact with the preheat zone of the flame. WithKa>10, these eddies can potentially interact with the reaction zone of the flame. In practice, however, eddies on the order of the Kolmogorov scale are dissipated quickly and do not interact with the flame and even are difficult to measure with PIV measurements. Therefore, in practice, eddies 10 times the size of will interact with the flame. t is estimated from the following equation: t = ( L 0 u U 03 ) 1 2 (6.6.4) u is the unburned mixture viscosity which is typically assumed to be based on the temperature at the jet exit.t f = f =S L is estimated from laminar flame calculations using the Premix code where f is the thermal thickness of the flame defined by: f = (T ad T u ) jrTj max (6.6.5) Now, turbulent Reynolds number which is the ratio of inertial forces in the flow to the viscous forces can be calculated by: Re t = U 0 L 0 u (6.6.6) Ka and Re t are plotted in Figure 6.18. Ka follows the trend u 0 along the centerline first increasing then decreasing.Re t increases linearly in proportion with the increase in integral length. 82 Figure 6.19: RescaledKa andRe t calculated at the burner centerline However, all the values are based on the unburnt viscosity u at the jet exit. Since temperature measurements at the burner centerline have also been measured, u can be calculated at each axial location based on the temperature of the cold jet propagating into ignited pilot and coflow. These values are then labeledKa rescaled andRe t;rescaled and they are plotted in Figure 6.19. With the additional effect of changing temperature at the centerline, both Ka and Re t have lower maximum values. Ka still follows the pattern ofu 0 , however,Re t stays fairly constant at a value of 500 along the centerline because of the increasing along the centerline which competes with the increasingL 0 . 6.6.2 Local Turbulent Flame Speed Measurements In addition toS T;GC described previously, an alternative approach to explore the ability of the PPJB to measure local turbulent flame speed has been developed. This method combines an averaged local velocity field obtained through PIV with an averaged 2-D flame surface. More precisely, the local turbulent flame speed S T,L is determined by projecting the local mean velocity vector onto the averaged flame surface at that location. This local turbulent flame speed may provide additional insight into the flame behavior at different axial locations in the flame as opposed to global measurements that produce a single value for a single flame. The c =0.5 isosurface is used as the average flame surface, as established in Section 6.4.3. The instantaneous velocity fields obtained from both solid and liquid particles are averaged to produce average velocity vector fields. The calculation ofS T,L uses a section of the flame front up to anx=D around 15 to avoid the flame tip, as it causes an exponential increase in the projection velocity. The flame edge is determined using the Sobel edge detection algorithm. The algorithm produces well- defined boundaries from an image of high gradient magnitude; the sharp cutoff in the binarized flame surface facilitates the generation of a well-defined edge contour. A sixth-order polynomial is then fit to the surface to allow for even grid spacing and gradients along the flame front. For each grid point on the flame surface, the closest velocity vector from the averaged velocity field is identified. Each grid point is then associated with a unit vector ^ n tangent to the flame surface. 83 Using ^ n, the local velocity vector is projected onto the flame surface to produceS T,L . S T,L = U local (U local ^ n) ^ n (6.6.7) This projection process is detailed in Figure 6.20 using two grid points. −0.8−0.7−0.6−0.5−0.4−0.3−0.2 10 11 12 13 14 15 16 r/D x/D U local S T,L ˆ n Figure 6.20: Sample Projection of Local Velocity onto Flame Surface Figure 6.21 illustrates the local turbulent speed along the brush surface as well as u 0 along the flame surface for various fuels at a constant S L of 30.3 cm/s. There is a 10% error in these measurements due to errors from the PIV data and the resolution of the flame surface. Figure 6.21:S T,L as a function ofx=D (left) andU 0 as a function ofx=D S T,L is plotted after the region of pilot influence, at x=D >5. S T,L has a monotonic increase withx=D. As the flame tip is approached,S T,L increases notably. This may be a result of flame wrinkle development downstream. The increase inS T,L is much higher for ethylene as the flame tip effects are captured at a lowerx=D than the other flames. The ratio of the velocity fluctuation over the laminar flame speed, u 0 =S L , is often used in correlations for the turbulent flame speed and in the Borghi diagram. [10, 88] Therefore, the local 84 u 0 =S L is now plotted along the averaged flame surface in order to explain the observed evolution ofS T,L . u 0 =S L decreases linearly along the flame surface by about a factor of 2. u 0 =S L also has a maximum value of 70 along the flame surface validating that the shear-generated turbulence can provide a fairly largeu 0 /S L close to the flame surface. S T,L from Abel-Inverted CH* Images (a) Methane Flame (b) Ethylene Flame (c) Jet A Flame Figure 6.22: Abel-Inverted CH* Images of Jet Flames with inner and outer CH* cutoff for projec- tion in the black line and maximum intensity represented by the dashed line As mentioned in an earlier section,S T,GC can be calculated from the global line-of-sight CH* surface. When the CH* layer is Abel-inverted, it can be visualized from a 2D perspective and more relevantly compared to the 2D PIV statistics. When looking at the Abel-inverted images, there is a clearly defined beginning and end to the CH* layer as a function of radial distance before the CH* layers merge at the flame tip. The inner and outer layer surfaces can be found by taking the cutoff of 0.5 times the maximum intensity at each radial position. This inner layer marks the start of the CH* reaction zone, so it should be a good indicator of the unburned vectors in the flame since CH* is a marker of reactivity. An additional CH* layer is defined as the location of maximum intensity at each radial location. An example of surfaces defined this way are shown for all three fuels in Figure 6.22. Now, the velocity is projected in the same manner onto the inner CH* surface and maximum CH* surface. The results are shown in Figure 6.23. Figure 6.23a) shows the inner layer which is the closest to the Mie scattering surface. A couple observations can be made: first, theS T,L increases exponentially toward the flame tip and has a much lower value than the Mie scattering value. S T,L shows some fuel-dependent differences with ethylene becoming the highest at the tip and methane and Jet A having fairly similar profiles. TheU 0 =S L at the flame surface still decreases, however, 85 (a) S T,L andU 0 projected onto CH* inner layer (b) S T,L andU 0 projected onto CH* max intensity surface Figure 6.23: S T,L as a function ofx=D (left) andU 0 as a function ofx=D for surfaces generated from Abel-inverted CH* it only decreases by around 30% instead of 50% like for the silicon oil surface. Moving outward, Figure 6.23b) showsS T,L calculated at the location of max CH* intensity at each radial location. S T,L has now an even lower maximum value which correlates to a dip in CH* chemiluminescence away from the influence of the pilot.S T,L then increases linearly after anx=D =10 toward the flame tip. Again, ethylene has the highest value ofS T,L while methane and Jet A are almost identical in value. Interesting,U 0 =S L at the location of maximum CH* is constant for all fuels. This indicates the presence of flame-generated turbulence along the CH* contour. S T,L has been defined for various averaged flame surfaces using an average PIV velocity field. It is apparent that the same differences between fuels can be seen in both the Mie scattering sur- face and the CH* projected surface and therefore in future studies, the CH* can be used for the projection. This parameter can provide an important benchmark for modeling studies to reproduce global flame behavior of different fuels in a statistically steady-state geometry. 86 6.6.3 Conditional Statistics As described in the Experimental Methods section, a novel edge detection technique based on solid particle edge detection has been developed using Mie scattering particle image density. Assuming an infinitely thin flame front due to the low resolution of these images, an instantaneous flame edge can be found based on sharp gradients in particle image intensity allowing a binary definition of burned vs. unburned vectors. This is possible since there is a noticeable drop in intensity after the flame due to local dilatation. This method is first tested on the full frame camera image where the flame thickness is negligible compared to the image resolution so the cutoff is easier to discern. Figure 6.24: Conditional Statistics ofU x andU 0 computed before the flame First the unburned side of the flame (Region U) is analyzed in Figure 6.24. The unburned layer has a noticeable peak in radial velocity all along the flame edge. The turbulence intensity also starts to increase slightly at the flame edge. The magnitude of the turbulence intensity however is still low. The mean axial velocity on the unburned side skews toward high values. V orticity is generated and peaks at the region just before the flame as well and decreases with increasing axial distance. Now examining the burned side of the flame (Region B) in Figure 6.25., the magnitudes of radial velocity and axial velocity are both lower in Region B due to enhanced mixing. u 0 peaks in the unburned layer as the thermal expansion from the flame no longer shields the central jet from mixing with the surrounding low velocity stream. Similarly, vorticity also increases in the burned region. Comparing the two figures, there is a noticeable increase in turbulence intensity from the un- burned to the burned side of the flame. This is due to the generation of turbulence intensity and thermal expansion from the flame and the lack of shielding from mixing with the surrounding coflow. The mean velocity also decreases after the flame. This jump in value indicates a marked 87 Figure 6.25: Conditional Statistics ofU x andU 0 computed after the flame difference between unburned and burned flowfield statistics. Figure 6.26: Probability Density Function (PDFs) of U x and U 0 computed before and after the flame Probability Density Functions (PDFs) of a region of the flame fromx=D =1114 were taken 88 to try to understand how the range of scales changes before and after the flame. The inert jet is shown in green and Jet A =0.6 and 0.8 are in blue and red respectively. Since there was no flame for the inert case, before and after the flame regions were defined as regions before and after the drop off in intensity due to the presence of the shear layer as a baseline for analysis. As described earlier, the velocity before the flame in Region A is higher than the velocity after the flame. In fact, there are no values below 50 m/s in the unburned region of the flame. The magnitude of velocity before the flame is lower for the inert jet than the ignited jet and a higher proportion of the energy is in smaller scales. After the flame, the peak burned velocity skews toward a higher velocity for the ignited jet and there are overall higher values of average velocity. This difference is attributed to the thermal expansion of the high velocity unburned vectors, which then mix and get dissipated in the shear layer. This indicates that the gradients in velocity in the shear layers are sharper due to the presence of the flame. The magnitude of turbulence intensity in Region A decreases with increasing reactivity. How- ever, the turbulence intensity in Region B is very similar for all cases even though the average values are very different. The conditional statistics demonstrate a marked difference in both aver- aged and r.m.s quantities across the flame surface. This type of analysis can be conducted for more zoomed in flame profiles in order to understand local density effects on the flame. 6.6.4 Nested PIV Although the stitched PIV can provide interesting statistics, it cannot resolve the smallest scales at the flame front and visualize how the vortices interact with the flame. A higher resolution window needs to be used for this purpose. However, when resolving the smallest scales, it becomes difficult to resolve the largest scales in the flow and capture the entire energy spectra of the flowfield. In order to capture the full flowfield statistics and understand how the largest scales interact with the smallest scales of the flow, a three-camera nested PIV configuration was tested. The first camera resolved the jet up to anx=D =14, while the smallest window resolved only around 1D of the jet centered aroundx=D =13. The resolution per pixel and the size of the smallest PIV interrogation window used for processing are listed in Table 6.2. This allows a simultaneously resolved range of 42 m per pixel to 3.8 m per pixel. Since the Kolmogorov scale of the unburnt mixture is calculated to be around 7m, this can resolve the smallest displacements in the flow. In order to visualize the velocity gradients, the cameras were sequentially focused on the shear layer of the flow. The 2D r.m.s. velocity plots for Jet A at =0.60 for each window are shown in Figure 6.27. It is clear that the magnitude ofu 0 increases with each successive window as the smaller scale fluctuations in velocity are being resolved. The 1D longitudinal energy spectra calculated for the flow is shown for Jet A at =0.60 in Figure 6.28. A wide range of wavenumbers were resolved. The falloff of the energy cascade as 89 Figure 6.27:Ux 0 profiles for Sequentially Zoomed Windows predicted by Pope was also reproduced [6]. No injection of energy was observed at the smaller scales for the range of wavenumbers investigated. As a disclaimer, in the derived energy spectra, no distinction can be made between the contribution of burned and unburned vectors, which is left for a future study. From the three different windows, it is clear that the PIV overestimates the magnitude of contribution of energy at the smaller scales for the full frame and the zoomed 1 windows. Figure 6.28:S T,L as a function ofx=D (left) andU 0 as a function ofx=D The Kolmogorov scale corresponded to a wavenumber k = 140, the Taylor microscale to k = 7, and L 0 to a k=0.3. The falloff of the energy spectra in the dissipation range is able to be captured by the zoomed 2 window. While these results are interesting, an edge detection or other 90 scalar measurement needs to be taken in order to determine the statistics and interaction with the flame surface. However, these results demonstrate that the full resolution of the range of relevant scales is possible. 6.7 OpenFOAM Turbulent Jet Flame Case Study The current diagnostics are limited to averaged velocity field, temperature measurements, and CH* but cannot directly measure heat release and other properties of direct relevance to the flame. To try to understand how other global flame observables may behave, a turbulent leann-dodecane/air flame at = 0.6 was modeled using OpenFOAM. While these results should not be directly compared to experiments since they are using RANS approximations, the correlation of species in comparison with heat release contours and flowfield statistics can be observed. Figure 6.29a) shows values ofu 0 on the left and U on the right, each scaled by their maximum values. Similar to experiments, the maximum turbulence intensity is generated in the shear layer. (a) Scaled u 0 (left) vs U x (right) (b) Scaled CH* (left) vs Heat Release (right) (c) Mean location of 523K contour (green) and CH* re- action layer (yellow) com- pared to mean Temperature profiles (right) Figure 6.29: Scaled Velocity, Temperature, and Chemiluminescence of Mean RANS Turbulent Jet Flame Also of interest was to understand how the CH* reaction layer compared with the mean 523 K temperature contour which was defined as the location where silicon oil will evaporate. As a reminder, both S T;GC and S T;L were determined based on these surfaces. A comparision be- tween temperature field contours and the CH* and 523 K disappearance contour is shown in Fig- ure 6.29b). On the left, the yellow contour represents the location of CH* and the green represents the average location of cold jet flow up to 523 K. Similar to experiments, the 523 K contour is 91 much shorter than the inner and outer CH* layers. The radial distancer=D of the 523 K contour also decreases steadily to the flame tip. The CH* contour on the other hand first increases then decreases in r=D from the jet centerline similar to experiments. An assumption made when vi- sualizing the global CH* contour was that it was a marker of flame reactivity. To investigate this assumption, location of the CH* contour was plotted versus heat release in Figure 6.29c). In- terestingly, the mean CH* layer is very close in thickness and location to the mean heat release, just slightly offset. The mean heat release decreases in maximum magnitude close to the flame tip and bottom of the burner where there is a higher mean velocity. However, the CH* is still present. Aside from this small difference, these results indicate that CH* is correlated to the global heat release contour. More detailed simulations should be performed to confirm these results for instan- taneous turbulent flame structure. The presence of excited species at the flame tip may indicate a mild combustion region at the flame tip with a distributed heat release zone. Of interest to future studies is understanding how well typical experimentally measured ob- servables correlate with temperature, velocity and CH* measurements. Figure 6.30a) shows mean contours of ethylene andn-dodecane. Asn-dodecane decomposes and is preheated, ethylene forms in a fairly thick layer before reacting. The results combined with heat release show that close to the location of heat release,n-dodecane is mostly decomposed into ethylene and othern-alkenes. The results are understandable given that the mean temperature of the unburned reactants is rising and n-dodecane generally pyrolyzes at temperatures above 500 C. This can indicate that in the presence of turbulence, the interaction of fuel fragments with the flame is more relevant than the interaction ofn-dodecane with the reaction layer. (a) Scaled ethylene (left) vs. n-Dodecane (right) (b) Scaled OH (left) vs. CH 2 O (right) (c) OHxCH 2 O (left) com- pared to dQ (right) Figure 6.30: Flame Markers in RANS Turbulent Jet Flame Figure 6.30b) shows scaled concentrations of OH and CH 2 O which are typically used as mark- ers of the reacted mixture and preheat zone respectively in experiments. CH 2 O has the same 92 broadened behavior as ethylene. OH is formed along the centerline at anx=D = 20 corresponding to the reacted region after the heat release layer. Figure 6.30c) shows the product of these two parameters which is often correlated to heat release layers in experiments. Interestingly, it follows a similar behavior compared to CH* in that there is the presence of this observable at the tip region of the flame where there is low heat release. Comparing this observable to the heat release profile, it matches almost precisely the spatial location of the heat release contour. These results further indicate the relevance of CH* as an experimental observable for global flame behavior. Formaldehyde is shown to be correlated with the pyrolysis and preheat zone of jet fuels. The product of OH and CH 2 O was also shown to be correlated with heat release as expected. However, these reaction markers are for statistically steady mean profiles of the flame shape. In- stantaneous LES snapshots should be taken to understand the correlation of these parameters with instantaneous flame behavior. The 523 K temperature contour showed a similar result to the silicon oil Mie Scattering contour and was correlated with the unburned reaction surface before pyrolysis ofn-dodecane. 6.8 Conclusions An extensive survey of different types of experimental observables for jet flames of heavy hydro- carbons has been made. In an attempt to understand the turbulent scales of interest as a function of axial distance along the centerline, full flame velocity statistics were captured using a three-camera stitched PIV apparatus with supplemental temperature measurements provided by a B-type ther- mocouple. Local turbulence properties like Karlovitz number and turbulent Reynolds number were seen to change along the centerline. This indicates that turbulence characteristics at the flame sur- face cannot be scaled by one value in the turbulent premixed regime diagram. PIV velocity fields show a very subtle fuel effect because the mean velocity profiles of all fuels were very similar. The maximum values of turbulence intensity and turbulent kinetic energy in the shear layer are only slightly higher for ethylene than Jet A and methane. Global consumption speed was calculated from Mie Scattering and Abel-Inverted CH* images and fuel effects were preserved in both mea- surements. However, the percentage difference in global consumption speed was smaller for CH* due to a larger value of flame height and a larger average flame area. Now, combining the two measurements, a local turbulent flame speed was defined by projecting the averaged velocity onto the averaged flame surface. This parameter shows similar fuel depen- dence and is consistently higher in value at the tip for ethylene compared to Jet A and methane. A new solid particle edge detection method was developed based on the sharp drop off in intensity before and after the flame. The statistics before and after the flame reveal a sudden increase in turbulence intensity in the vectors after the flame and also a decrease in mean velocity. This shows 93 that there is a difference in mean flow properties due to the presence of the flame. A novel nested PIV configuration was tested with three cameras and was found to be able to resolve the full energy spectra simultaneously. With the addition of a scalar tracer in the flow, these nested PIV images can provide a powerful tool for understanding the interaction of the flame surface with small and large-scale structures in the flow. Finally, RANS simulations of the jet burner showed similarities in location of global flame observables and show a correlation of CH* with heat release. 94 Chapter 7 Conclusions and Recommendations 7.1 Concluding Remarks A Piloted Premixed Jet Burner was developed at the University of Southern California to investi- gate the behavior of heavy hydrocarbons in the presence of highly turbulent premixed jet flames. In order to understand how heavy hydrocarbon fuels may behave when exposed to eddies in the preheat zone of the flame, mass burning rates and extinction strain rates were computed in a coun- terflow flame apparatus for varying partially-decomposed mixtures ofn-dodecane. It was found that asn-dodecane decomposes, it formsn-alkenes, H 2 and methane which then interact directly with the flame front. Mass burning rates of partially reacted mixtures changed only by around 15% of the initial value while extinction strain rates increased by up 150% This was due to the increased presence of ethylene . The values of extinction strain rates were found to be sensitive to the dif- ferential diffusion of C 2 H 4 -N 2 in the flame front for lean flames. This indicates that fuel pyrolysis from mixing in the preheat zone due to turbulence can cause major differences in fundamental flame properties. With these results in mind, a thorough investigation of various lean hydrocarbon/air flames at differing , jet Reynolds number, and degree of adiabaticity of the coflow was conducted in the piloted premixed jet burner. To provide relevant data for modelers, the boundary conditions of temperature and velocity at the burner exit were characterized in addition to the 2D velocity fields of the flame. Temperatures were also measured along the centerline. These results were able to be reproduced by LES simulations in collaboration with the California Insititute of Technology. This database of values showed that local non-dimensional Karlovitz number and turbulent Reynolds number vary with axial distance along the burner centerline. Line-of-sight images of chemiluminescence provided a global metric for overall reactivity of the jet flame. A couple key observables were defined from the chemiluminescence images. Flame height was defined from line-of-sight centerline chemiluminescence profiles. Flame height de- creases with increasing global turbulent consumption speed defined from the area of global CH* flame shape. Flame height for all fuels was found to scale with laminar flame speed. However, even with adiabatic conditions in the coflow, ethylene had the shortest flame height while methane had the longest flame height. The rest of the fuels are in between these two extremes for all Reynolds numbers and are all very close in value to one another. At higher Reynolds numbers this 95 discrepency increases and these results are able to be reproduced by LES modeling of the burner. Mie scattering surfaces also confirm these fuel-dependent findings. Abel-Inverted images of global flame shape allowed for two separate definitions of flame surface. The first was dependent on the start of the CH* reaction layer and the second was defined as the location of maximum intensity of CH* as a function of radial distance. These results were used to define a flame surface upon which the 2D average velocity field can be projected. This value was defined as the local turbulent flame speed and is an additional benchmark for modelers to reproduce. Fuel-dependent effects are also seen for these measurements particularly close to the flame tip. Several types of novel Particle Image Velocimetry measurements specific to the highly tur- bulent jet flame apparatus were developed. To charaterize the statistics of the entire flowfield at a higher resolution, three cameras were stitched together. These velocity field measurements showed very little differences between fuels; however, they provided accurate characterization of integral length of the flowfield. In order to resolve the vectors before and after the flame, an edge detection technique based on local particle intensity gradients was developed. While this type of analysis is common for laminar flames, it has not been attempted in this configuration before. Results showed a sharp increase in turbulence intensity in the vectors after the flame. This analysis can also be extended to determine a local density based on particle density at each location and potentially have simultaneous density and velocity fields within the flame. Finally, in order to visualize the small eddies that are interacting with the flow, a nested PIV setup was developed. Using three cameras and one laser, this technique was able to effectively capture the entire energy spectra of the flowfield. Although the current spectra included both burned and unburned vectors, vortices on the order of the Kolmogorov scale can be visualized with this technique. Coupled with the addition of PLIF measurements to determine the location of the flame, the energy spectra can be defined separately before and after the flame. Local interactions between vortices and the flame surface can be achieved. 7.2 Recommendations for Future Work While some fuel-dependent effects were found based on the global flame behavior, further work should involve time-resolved data in order to investigate stochastic effects and their influence on local flame behavior. Important to the development of this setup is the addition of formaldehyde and OH PLIF to visualize instantaneous flame structure and instantaneous heat release in the flame front. The development of additional scalar measurements coupled with the PIV system will al- low a clear demarcation between burned and unburned vectors and the local strain rate can be calculated. Additionally, time-resolved PIV measurements should be made in order to attempt to identify vortices that are stretching the flame front and see if the local vorticity at the flame surface 96 is different for different fuel/air mixtures. Furthermore, the density detection algorithm used for determination of burned versus unburned vectors can be modified for PIV windows with better resolution to define an instantaneous particle density at each location in the image. This particle density can be correlated to a local temperature or gas density of the jet. This will enable the determination of flame thickness and differences in local density gradients at each axial location due to the presence of the flame. The nested PIV images may be used in this approach. Finally, additional OpenFOAM simulations can be performed to resolved higher exit velocities and with more complex chemical mechanisms. Instead of a steady-state approach, LES of the burner would be useful to determine burned versus unburned statstics before and after the flame. Instantaneous snapshots of the interactions of velocity vectors and scalars within the flame front can show key difference between n-dodecane/air flames and methane/air flames. These results could complement the formaldehyde PLIF and OH PLIF measurements being developed. 97 References [1] T. I. E. Agency. [2] R. Royce, The Jet Engine, Rolls-Royce PLC, 1986. [3] C. K. Law, Cambridge University Press. [4] Y . Gao, R. Shan, S. Lyra, C. Li, H. Wang, J. H. Chen, T. Lu, Combust. Flame 163 (2016) 427–436. [5] A. N. Kolmogorov, Proceedings of the USSR Academy of Sciences 30 (1941) 299–303. [6] S. Pope, Turbulent Flows, Cambridge University Press, 2000. [7] N. R. Panchapakesan, J. L. Lumley, J. Fluid Mech. 246 197–223. [8] N. Peters, Turbulent Combustion, Cambridge University Press, 2000. [9] N. Peters, J. Fluid Mech. 384 (1999) 107 – 132. [10] J. Duclos, D. Veynante, Recent Adv. Aero. Sci. (1985) 117 – 134. [11] N. Peters, Proc. Comb. Inst. 21 (1986) 1231 – 1250. [12] G. Damkohler, Z. Elektrochem 46 (1940) 601 – 652. [13] J. Driscoll, Prog. Energ. Combust. 34 (1) (2008) 91 – 134. [14] A. Lipatnikov, J. Chomiak, Prog. Energ. Combust. 36 (1) (2010) 1 – 102. [15] C. Ji, E. Dames, Y . Wang, H. Wang, F. Egolfopoulos, Combust. Flame 157 (2010) 277–287. [16] S. G. Davis, A. V . Joshi, H. Wang, F. Egolfopoulos, Proc. Comb. Inst. 30 (1) (2005) 1283 – 1292. [17] X. You, F. Egolfopoulos, H. Wang, Proc. Combust. Inst 32 (2009) 403–410. [18] A. Holley, X. You, F. Egolfopoulos, H. Wang, Proc. Combust. Inst 32 (2009) 1157–1163. [19] D. Bradley, M. Lawes, K. Liu, R. Woolley, Proceedings of the Combustion Institute 31 (1) (2007) 1393 – 1400. 98 [20] M. Lawes, M. P. Ormsby, C. G. Sheppard, R. Woolley, Combustion and Flame 159 (5) (2012) 1949 – 1959. [21] C. Mouna¨ ım-Rousselle, L. Landry, F. Halter, F. Foucher, Proceedings of the Combustion Institute 34 (2) (2013) 2941 – 2949. [22] P. Aleiferis, J. Serras-Pereira, D. Richardson, Fuel 109 (2013) 256 – 278. [23] K. Goh, P. Geipel, F. Hampp, R. Lindstedt, Proc. Combust. Inst. 34 (2013) 3311–3318. [24] F. Dinkelacker, A. Soika, D. M. et al., Proc. Combust. Inst 27 (1998) 857–865. [25] A. Poludnenko, E. Oran, Combust. Flame 157 (2010) 995 – 1011. [26] Y .-C. Chen, R. Bilger, Combust. Flame 131 (2002) 400 – 435. [27] M. Dunn, A. Masri, R. Bilger, Combust. Flame 151 (2007) 46 – 60. [28] M. Dunn, A. Masri., R. Bilger, R. Barlow, G.-H. Wang, Proc. Combust. Inst. 32 (2009) 1779 – 1786. [29] A. Lipatnikov, J. Chomiak, Prog. Energ. Combust. 31 (1) (2005) 1 – 73. [30] H. Kido, T. Kitagawa, K. Nakashima, K. Kato, Mem. Fac. Eng, Kyushu Univ. 49 (1989) 229–247. [31] M. Nakahara, H. Kido, Mem. Fac. Eng, Kyushu Univ. 58 (1998) 55–82. [32] V . Karpov, E. Severin, Combust. Explos. Shock Waves 16 (1980) 41–46. [33] M. Mansour, N. Peters, Y . Chen, Proc. Combust. Inst 27 (1998) 763–773. [34] M. Mansour, N. Peters, Y . Chen, Proc. Combust. Inst 27 (1998) 811–818. [35] P. H. R. et al., Prog. Energ. Combust. 26 (2000) 225–282. [36] Y .-C. Chen, R. Bilger, Combust. Flame 138 (2004) 155 – 174. [37] J. M. Samaniego, F. N. Egolfopoulos, C. T. Bowman, Combust. Sci. Technol. 109 (1995) 183–203. [38] H. N. Najm, P. H. Paul, C. J. Mueller, P. S. Wyckoff, Combust. Flame 113 (312-332). [39] T. Kathrotia, U. Riedel, A. Seipel, K. Moshammer, A. Brockhinke, Appl. Phys. B. 107 (2012) 571–584. 99 [40] F. Carbone, J. L. Smolke, A. M. Fincham, F. N. Egolfopoulos, Combust. Flame. [41] A. Melling, Meas. Sci. Technol. 8 (1997) 1406–1416. [42] A. T. Hjemfelt, L. F. Mockros, Appl. Sci. Res 16 (1996) 149. [43] A. Stella, G. Guj, J. Kompenhaus, M. Raffel, H. Richard, Exp. Fluids (2001) 167–180. [44] R. Schl¨ ußler, J. Czarske, A. Fischer, Optics and Lasers in Engineering 54 (2014) 93–104. [45] A. Fincham, G. Spedding, Exp. Fluids 23 (1997) 449–462. [46] R. Kee, J. Grcar, M. Smooke, J. Miller, Sandia Report (SAND85-8240). [47] R. Kee, F. Rupley, J. Miller, Sandia Report (SAND89-8009). [48] R. Kee, J. Warnatz, J. Miller, Sandia Report (SAND83-8209). [49] H. Wang, E. Dames, B. Sirjean, D. A. Sheen, R. Tango, A. Violi, J. Y . W. Lai, F. N. Egol- fopoulos, D. F. Davidson, R. K. Hanson, C. T. Bowman, C. K. Law, W. Tsang, N. P. Cernan- sky, D. L. Miller, R. P. Lindsted. [50] A. Lutz, R. Kee, J. Miller, Report SAND87-8248. [51] F. N. Egolfopoulos, Proc. Combust. Inst. 25 (1994) 1375 – 1381. [52] S. Lapointe, B. Savard, G. Blanquart, Combust. Flame 162 (2015) 3341–3355. [53] B. Savard, G. Blanquart, Combust. Flame. [54] OpenFOAM. URLwww.openfoam.org [55] k-omega Shear Stress Transport (SST). URLhttp://www.openfoam.com/documentation/cpp-guide/html/guide-turbulence-ras-k-omega-sst.html [56] A. Vie, B. Franzelli, Y . Gao, T. Lu, H. Wang, M. Ihme, Proc. Combust. Inst 35 (2015) 1675– 1683. [57] N. Peters, V ol. 32, 2009, pp. 1–25. [58] H. Wang, R. Xu, K. Wang, C. Bowman, R. Hanson, D. Davidson, F. Egolfopoulos, Combust. Flame. [59] C. L. S. Davis, Combust. Sci. Technol. 140 (1998) 427–449. 100 [60] M. St¨ ohr, I. Boxx, C. Carter, W. Meier, Combust. Flame 157 (2012) 2636–2649. [61] C. Ji, Y . Wang, F. Egolfopoulos, J. Prop. Power 27 (2011) 856. [62] Y . Wang, A. H. C. Ji, F. Egolfopoulos, T. Tsotsis, H. Curran, Proc. Combust. Inst. 32 (2009) 1035–1042. [63] C. Law, C. Sung, H. Wang, T. Lu, AIAA Journal 41 (2003) 1629–1643. [64] F. A. Williams, Combustion Theory, Benjamin/Cummings, Menlo Park, CA, 1985. [65] K. Bray, R. Cant, Proc. R. Soc. London A 434 (1890) (1991) 217 – 227. [66] C. Westbrook, W. Pitz, O. Herbinet, H. Curran, E. Silke, Combust. Flame 156 (2009) 181– 199. [67] P. Dagaut, A. E. Bakali, A. Ristori, Fuel 85 (2006) 944–956. [68] K. Dahm, P. Virk, R. Bounaceur, F. Battin-Leclerc, P. Marquaire, R. Fournet, E. Daniau, M. Bouchez, Journal of Analytical and Applied Pyrolysis 71 (1004) 865–881. [69] B. Sirjean, E. Dames, D. Sheen, X. You, C. Sung, A. Holley, F. Egolfopoulos, H. Wang, S. Vasu, D. Davidson, online. [70] H. Wang, X. You, A. Joshi, S. Davis, A. Laskin, F. Egolfopoulos, C. Law, online. [71] A. Lutz, R. Kee, J. Miller, SENKIN: A FORTRAN program for predicting homogeneous gas phase chemical kinetics with sensitivity analysis, Tech. Rep. SAND88-8248, Sandia National Laboratories, Alburquerque, New Mexico (1988). [72] A. Lutz, R. Kee, J. Grcar, F. Rupley, OPPDIF: A Fortran program for computing opposed- flow diffusion flames, Tech. Rep. SAND96-8243, Sandia National Laboratories, Albur- querque, New Mexico (1997). [73] F. N. Egolfopoulos, Proc. Combust. Inst. 25 (1994) 1375–1381. [74] T. Malewicki, K. Brezinsky, Proc. Combust. Inst 34 (2012) 361–368. [75] D. Davidson, Z. Hong, G. Pilla, A. Farooq, R. Cook, R. Hanson, Combust. Flame 157 (2010) 1899–1905. [76] D. Davidson, Z. Hong, G. Pilla, A. Farooq, R. Cook, R. Hanson, Proc. Combust. Inst. 33 (2011) 151–157. 101 [77] C. K. Law, Combustion Physics, Cambridge University Press, 2010. [78] M. Colket, T. Edwards, S. Williams, N. Cernansky, D. Miller, F. Egolfopoulos, P. Lindstedt, K. Seshadri, F. Dryer, C. Law, D. Friend, D. Lenhert, H. Pitsch, A. Sarofim, M. Smooke, W. Tsang, in: 45th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, Jan. 8-11, 2007. [79] T. Edwards, J. Prop. Power 19 (6) (1994) 1089 – 1107. [80] H. Pitsch, Combust. Flame 143 (4) (2005) 587 – 598. [81] J. A. van Oijen, F. A. Lammers, L. P. H. de Goey, Combust. Flame 127 (3) (2001) 2124 – 2134. [82] O. Gicquel, N. Darabiha, D. Thevenin, Proc. Comb. Inst. 28 (2) (2000) 1901 – 1908. [83] A. Aspden, M. Day, J. Bell, J. Fluid Mech. 680 (2011) 287 – 320. [84] A. Aspden, M. Day, J. Bell, Proc. Comb. Inst. 35 (2015) 1321–1329. [85] A. Aspden, M. Day, J. Bell, Proc. Comb. Inst. 33 (2011) 1473 – 1480. [86] R. Sankaran, E. Hawkes, C. Yoo, J. Chen, Combust. Flame 162 (2015) 3294 – 3306. [87] J. Smolke, S. Lapointe, L. Paxton, G. Blanquart, F. Carbone, A. M. Fincham, F. N. Egol- fopoulos, Proc. Combust. Inst 36 (2017) 1877–1884. [88] R. Abdel-Gayed, D. Bradley, Philos. T. Roy. Soc. A 301 (1457) (1981) 1 – 25. [89] S. Filatyev, J.F., C. Carter, J. Donbar, Combust. Flame 141 (1) (2005) 1 – 21. [90] J. Duclos, D. Veynante, Combust. Flame 95 (1) (1993) 101 – 117. [91] A. Lipatnikov, J. Chomiak, Prog. Energ. Combust. 28 (1) (2002) 1 – 74. [92] I. Shepherd, Proc. Combust. Inst. 26 (1996) 373 – 379. [93] P. Cho, C. Law, J. Hertzberg, R. Cheng, Proc. Combust. Inst. 21 (1986) 1493 – 1499. [94] I. Shepherd, R. Cheng, Combust. Flame 127 (3) (2001) 2066 – 2075. [95] P. Tamadonfar, O. Gulder, Combust. Flame 162 (1) (2015) 115 – 128. [96] M. Dunn, A. Masri, R. Bilger, Combust. Flame 151 (1) (2007) 46 – 60. 102 [97] J. Sidey, E. Mastorakos, Proc. Combust. Inst. 35 (2014) 3537–3545. [98] T. Lee, G. North, D. Santavicca, Combust. Flame 93 (4) (1993) 445 – 456. [99] C. M. de Silva, C. Atkinson, N. A. Buchmann, E. Gnanamanickam, N. Hutchins, J. Soria, I. Marusic, 10th International Symposium on Particle Image Velocimetry. 103
Abstract (if available)
Abstract
An investigation of fuel and hydrodynamic effects is performed on piloted premixed jet flames. The investigation is carried out at varying laminar flame speed, varying heat losses, jet Reynolds number, fuel molecular weight, and fuel chemical classification. Well-characterized boundary conditions, well-resolved two-dimensional velocity fields from particle image velocimetry, and line-of-sight CH* profiles are analyzed. Global flame reactivity was analyzed from CH* and Mie Scattering images to determine global and local flame speed. ❧ Experimental results indicate that small amounts of heat losses may play a significant role on the jet reactivity as the flame heights scale with the heat loss from the jet. However, differences between flames with different fuels can still be seen in the absence of heat losses and these differences are magnified at higher Reynolds numbers. Particularly, methane flames have consistently smaller global consumption speeds than ethylene for a given laminar flame speed and other fuels present approximately the same flame height at an intermediate value between the two extremes. ❧ A full characterization of velocity flowfield of the jet was performed using multiple Particle Image Velocimetry experiments. Highly resolved single camera PIV at the exit plane was able to characterize the boundary conditions at the exit of the jet, while stitched PIV allowed for the characterization of local turbulence properties interacting with the flame. The nested PIV configuration was used to resolve the entire energy spectra of vectors interacting with the flame by sequentially focusing in on smaller viewing windows until the Kolmogorov scale was resolved. Finally, edge detection techniques were developed for conditioned burned and unburned statistics. All of these measurements using fuels from methane to Jet A provide an experimental database of results that can be used as benchmarks for modeling studies of highly turbulent premixed flows.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Studies of methane counterflow flames at low pressures
PDF
Investigations of fuel and hydrodynamic effects in highly turbulent premixed jet flames
PDF
Experimental and kinetic modeling studies of flames of H₂, CO, and C₁-C₄ hydrocarbons
PDF
Studies on the flame dynamics and kinetics of alcohols and liquid hydrocarbon fuels
PDF
CFD design of jet-stirred chambers for turbulent flame and chemical kinetics experiments
PDF
Studies of combustion characteristics of heavy hydrocarbons in simple and complex flows
PDF
Pressure effects on C₁-C₂ hydrocarbon laminar flames
PDF
Experimental studies of high pressure combustion using spherically expanding flames
PDF
Experimental investigation of the propagation and extinction of edge-flames
PDF
Flame ignition studies of conventional and alternative jet fuels and surrogate components
PDF
Flame characteristics in quasi-2D channels: stability, rates and scaling
PDF
Modeling investigations of fundamental combustion phenomena in low-dimensional configurations
PDF
Determination of laminar flame speeds under engine relevant conditions
PDF
Re-assessing local structures of turbulent flames via vortex-flame interaction
PDF
Direct numerical simulation of mixing and combustion under canonical shock turbulence interaction
PDF
Accuracy and feasibility of combustion studies under engine relevant conditions
PDF
End-gas autoignition investigations using confined spherically expanding flames
PDF
Studies of siloxane decomposition in biomethane combustion
PDF
Development of a novel heterogeneous flow reactor: soot formation and nanoparticle catalysis
PDF
Studies of Swiss-roll combustors for incineration and reforming applications
Asset Metadata
Creator
Smolke, Jennifer
(author)
Core Title
Investigations of fuel effects on turbulent premixed jet flames
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Mechanical Engineering
Publication Date
10/21/2019
Defense Date
08/30/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
decomposition,hydrocarbon fuel,jet fuel,laminar flames,OAI-PMH Harvest,premixed flames,turbulent flames
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Egolfopoulos, Fokion (
committee chair
), Ronney, Paul (
committee member
), Shing, Katherine (
committee member
), Spedding, Geoffrey (
committee member
)
Creator Email
smolke@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-444977
Unique identifier
UC11265575
Identifier
etd-SmolkeJenn-5807.pdf (filename),usctheses-c40-444977 (legacy record id)
Legacy Identifier
etd-SmolkeJenn-5807.pdf
Dmrecord
444977
Document Type
Dissertation
Rights
Smolke, Jennifer
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
decomposition
hydrocarbon fuel
jet fuel
laminar flames
premixed flames
turbulent flames