Design and implementation of a novel, fuel-flexible, laboratory-scale, premixed combustor

Permanent Link: http://ufdc.ufl.edu/UFE0041339/00001

Material Information

Title: Design and implementation of a novel, fuel-flexible, laboratory-scale, premixed combustor
Physical Description: 1 online resource (155 p.)
Language: english
Creator: Fond, Benoit
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009


Subjects / Keywords: blowout, combustion, diagnostics, flame, flameless, flexible, laboratory, optical, premixed, reactor, scale, stirred, turbulent
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Mechanical Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Abstract: Design and implementation of a novel, fuel-flexible, laboratory-scale, premixed combustor Recent developments in the production of fuels from alternative sources such as biomass raise a global interest in burning these highly variable fuels in gas turbine combustors. However, current low emission burner technologies and combustor stability are very sensitive to fuel composition.The current study aims at designing a fuel flexible combustion system which can be used for kinetic investigations, as well as a future candidate for low purity fuels combustion. A novel premixed burner design has been created based on a concept of a forced recirculation by jet impingement and a strong stirring. A numerical investigation of the combustion performance has been performed on a fully detailed model to define an operative range of interest. The entire system has been designed to provide this range of operation based on detailed heat transfer cooling and gas dynamics injector models. Optical and physical diagnostics capabilities have been implemented in order to provide insight into the reactor chemistry. This will allow the characterization of the combustor performance and kinetic investigation of alternative fuel combustion on the one hand, and of the kinetic effects of CO2 and H2O in flameless combustion on the other hand.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Benoit Fond.
Thesis: Thesis (M.S.)--University of Florida, 2009.
Local: Adviser: Hahn, David W.
Local: Co-adviser: Lear, William E.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041339:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041339/00001

Material Information

Title: Design and implementation of a novel, fuel-flexible, laboratory-scale, premixed combustor
Physical Description: 1 online resource (155 p.)
Language: english
Creator: Fond, Benoit
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009


Subjects / Keywords: blowout, combustion, diagnostics, flame, flameless, flexible, laboratory, optical, premixed, reactor, scale, stirred, turbulent
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Mechanical Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Abstract: Design and implementation of a novel, fuel-flexible, laboratory-scale, premixed combustor Recent developments in the production of fuels from alternative sources such as biomass raise a global interest in burning these highly variable fuels in gas turbine combustors. However, current low emission burner technologies and combustor stability are very sensitive to fuel composition.The current study aims at designing a fuel flexible combustion system which can be used for kinetic investigations, as well as a future candidate for low purity fuels combustion. A novel premixed burner design has been created based on a concept of a forced recirculation by jet impingement and a strong stirring. A numerical investigation of the combustion performance has been performed on a fully detailed model to define an operative range of interest. The entire system has been designed to provide this range of operation based on detailed heat transfer cooling and gas dynamics injector models. Optical and physical diagnostics capabilities have been implemented in order to provide insight into the reactor chemistry. This will allow the characterization of the combustor performance and kinetic investigation of alternative fuel combustion on the one hand, and of the kinetic effects of CO2 and H2O in flameless combustion on the other hand.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Benoit Fond.
Thesis: Thesis (M.S.)--University of Florida, 2009.
Local: Adviser: Hahn, David W.
Local: Co-adviser: Lear, William E.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041339:00001

This item has the following downloads:

Full Text




2 2009 Benoit Fond


3 To my parents and my sisters


4 ACKNOWLEDGMENTS I am very grateful to Dr. David Hahn for his constant support and advisement throughout my masters degree at th e University of Florida His scientific knowledge and intellectual curiosity have been a very valuable inspiration for me. In addition, Dr Hahn has always shown an unlimited patience and provided me encouragement. I would also like to thank Mark Harris for sharing his great experience in the design of gas turbine combustors. I acknowledge my sincere thanks to Dr William Lear for serving as a cochair in my committee and to Dr Nicolo Omenetto and Dr David Mikolaitis for their valuable teaching and advice. I am also thankful to my labmates (Bret, Prasoon, Phil, Michael A, Michael B, Richard and Sarah) to their contribution in making the laboratory a very pleasant place for research and discussion. Finally, I would like to thank my parents and my five sisters for their constant encouragement in pursuing my studies across the Atlantic.


5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...................................................................................................... 4 LIST OF TABLES ................................................................................................................ 7 LIST OF FIGURES .............................................................................................................. 8 LIST OF ABBREVIATIONS .............................................................................................. 10 ABSTRACT ........................................................................................................................ 11 CHAPTER 1 INTRODUCTION ........................................................................................................ 13 2 STATEMENT OF SCOPE .......................................................................................... 20 Objectives o f t he Combustion Rig .............................................................................. 20 Reaction Limited Operation. ................................................................................ 20 Stability ................................................................................................................. 21 Control a nd Diagnostics ....................................................................................... 25 Fuel a nd Gases .................................................................................................... 28 Constraints a nd Limitations ........................................................................................ 32 Flow Rates ........................................................................................................... 32 Pressure ............................................................................................................... 33 Volume ................................................................................................................. 35 Inlet a nd Reactor Temperature ............................................................................ 35 3 DESIGN O F THE BURNER ....................................................................................... 42 System Approach ....................................................................................................... 42 Combustor Geometry ................................................................................................. 43 Kinetic Modeling ......................................................................................................... 47 Heat Transfer Model ................................................................................................... 61 Gas Dynamics Model ................................................................................................. 77 4 FINAL DIMENSIONS OF THE RIG ........................................................................... 84 5 DESIGN OF THE DELIVERY SYSTEM .................................................................... 92 A ir Supply .................................................................................................................... 92 Compressor, Dryer, Filters a nd Receiving Tank ................................................. 92 Air Heaters ........................................................................................................... 95 Fuel a nd Air Mixing System ........................................................................................ 95 Control o ver The Gas Lines ..................................................................................... 105


6 6 DIAGNOSTICS AND FUTURE IMPLEMENTATIONS ............................................ 112 Combustor Characterization ..................................................................................... 113 Species a nd Temperature Characterization i n t he Primary Zone ........................... 114 Soot Characterization ............................................................................................... 123 Futures Studies a nd Implementations: .................................................................... 127 Conclusions .............................................................................................................. 131 APPENDIX A EQUATIONS OF THE GAS DYNAMICS MODEL ................................................... 132 B MACHINING DRAWINGS ........................................................................................ 134 LIST OF RE FERENCES ................................................................................................. 150 BIOGRAPHICAL SKETCH .............................................................................................. 155


7 LIST OF TABLES Table page 2 -1 Some physical and chemical properties of the different fuels involved. ............... 41 3 -1 Combustion air flow rates ...................................................................................... 60 3 -2 Input parameters of the heat trans fer model ......................................................... 62 3 -3 Performance functions for material choices .......................................................... 71 3 -4 Properties and composition of some tentative materials ...................................... 74 4 -1 List of the main parts of the combustion rig .......................................................... 84 4 -2 Machining parameters used ................................................................................... 91 5 -1 List of products for the air supply ........................................................................... 94 5 -2 Range of total air and fuel flow rates ..................................................................... 96 5 -3 Non-dimensional n umbers value for the creeping flow discussion .................... 100 5 -4 Parameters for the mixing system ....................................................................... 104 5 -5 List of material used in the air/ fuel mixing system .............................................. 104 5 -6 Range of gases and control configurations ......................................................... 106 5 -7 List of flow controllers ........................................................................................... 109 5 -8 Flow rates of H2, CO and global combustible mixture (air and fuel) for stoichiometric mixtures of CO/H2/Air containing from 5 to 50% of H2 in mass of fuel. ................................................................................................................... 111 6 -1 Selected experim ental studies on soot formation ............................................... 125


8 LIST OF FIGURES Figure page 2 -1 Heat balance in a Well Stirred reactor ................................................................... 23 2 -2 Absolute pressure evolutions (in kPa) along the flow path of a combustion system burning at negative pressure A) and a positive pressure B). ................... 35 2 -3 Adiabatic flame temperature for mixtures of CH4/O2/N2 ....................................... 37 2 -4 Composition at equilibrium for different methane -air mixtures at initial temperature of 400K and a constant pr essure of 1 atmosphere. ......................... 37 3 -1 Combustion System approach for premixed liquid fuel/air mixture ...................... 42 3 -2 Normal A) and axial B) cut of the burner ............................................................... 44 3 -3 CAD view of the primary zone of the burner ......................................................... 45 3 -4 Flexibility in the injector design .............................................................................. 45 3 -5 Primary (red), secondary (green) and dilution zone (blue) used in the reactor network model A) and flow transfer between WSR according to the model 1 B). ........................................................................................................................... 50 3 -6 Temperature and residence time obtained at LP=1 (model 1) ............................. 51 3 -7 Temperature and residence time obtained at LP=2 (model 1) ............................. 53 3 -8 Exit temperature and residence time in the primary reactor for two different loadings simulated using the model 1). ................................................................. 55 3 -9 Temperature evolution in a WSR for a given composition ................................... 55 3 -10 Temperature and residence time obtained with computer simulations (model 2) ............................................................................................................................. 58 3 -11 Flow transfer between CSTRs a ccording to the model 2. .................................... 58 3 -12 Representation of the cooling passages ............................................................... 67 3 -13 Influence of the conductivity in the heat transf er, experimental results (A) and steady state wall model (B) .................................................................................... 70 3 -14 Thermal properties of different materials ............................................................... 74 3 -15 Cooling f lows required A) and axial temperature changes B) .............................. 76


9 3 -16 Gas dynamics model programming scheme ......................................................... 78 3 -17 Pressure and Mach number evolution in the combustion air flow path ................ 82 3 -18 Pressure drop across the injector for different injectors ....................................... 82 3 -19 Jet Mach nu mber for different injectors ................................................................. 83 4 -1 Axial view of the rig (Scale1/1) .............................................................................. 85 4 -2 Normal view of the rig (resized) ............................................................................. 86 4 -3 Rig assembled, outside A), and combustion zone B) ........................................... 88 4 -4 Back view of the rig with distribution plate A) and without B) ............................... 88 4 -5 Injector design: previous invasive design A) and final design with normal B) and C) isometric views ........................................................................................... 89 4 -6 Injector ports on rig casing and l iner A) and injector B) ........................................ 90 5 -1 Pressure evolution in the tank at 35 SCFM. .......................................................... 94 5 -2 Mixing system ....................................................................................................... 101 5 -3 Air supply Process diagram ................................................................................. 107 5 -4 Process diagram of liquid fuel .............................................................................. 107 5 -5 Process diagram s of gas in level 4 A), 2 B) and 1 C) ......................................... 108 6 -1 Picture of the thermocouple ports A) and an optical access port B) .................. 112 6 -2 Spheric al coordinate system in elastic light scattering ....................................... 118 6 -3 Staged combustion configuration where the two rich injectors are shown ........ 1 28 6 -4 Different geometric modifications: removal and change of the inner tube A), addition of 8 secondary air holes B). ................................................................... 129 6 -5 Additional UV window port for 90 optical access. .............................................. 130 B 1 Machining drawings ............................................................................................. 135


10 LIST OF ABBREVIATION S NOx n itrogen o xides PAH polycyclic aromatic hydrocarbons A s ection (m2) Cp s pecific heat at constant pressure (J.kg1. K) k t hermal conductivity (W.m1.K1) m m ass flow rate (kg.s1) M M ach number P p ressure (Pa) Q heat (J) R u niversal gas constant (8.314 J.K1. mol1) T t emperature (K) V v olume (m3) a bsorptivity e missivity w avelength (m) d ynamic v iscosity (Pa.s1) d ensity (kg.m3) e quivalence ratio g g as (subscript) i i nlet (subscript) o s tagnation condition (subscript) l liquid (subscript) w w all (subscript) s onic condition (subscript)


11 Abstract of Thesis Presented to the Graduate S chool of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DESIGN AND IMPLEMENTATION OF A NOVEL FUEL -FLEXIBLE, LABORATORY SCALE PREMIXED COMBUSTOR By Benoit Fond December 2009 Chai r: David W. Hahn Cochair: William E. Lear Major: Mechanical Engineering Recent developments in the production of fuels from alternative sources such as biomass raise a global interest in burning these highly variable fuels in gas turbine combustors. However, current low emission burner technologies and combustor stability are very sensitive to fuel composition. The current study aims at designing a fuel flexible combustion system which can be used for kinetic investigations, as well as a future candidate for low purity fuels combustion. A novel premixed burner design has been created based on a concept of a forced recirculation by jet impingement and a strong stirring. A numerical investigation of the combustion performance has been performed on a fully detailed model to define an operative range of interest. The entire system has been designed to provide this range of operation based on detailed heat transfer cooling and gas dynamics injector models. Optical and physical diagnostics capabilities have been implemented in order to provide insight into the reactor chemistry. This will allow the characterization of the combustor performance and kinetic investigation of alternative fuel combustion on the


12 one hand, and of the kinetic effects of CO2 and H2O in flameless combustion on the other hand.


13 CHAPTER 1 INTRODUCTION Environmental awareness has been a leading force for gas turbine development in the recent years. The economic value of energy no longer includes only the processing cost from the fuel source to the fuel but attributes a real part to pollutant emission upon power generation. This latter part could take different forms s uch as the carbon tax on CO2 emissions in Scandinavian countries, or the emission trading with the cap and trade policy in the United States. Since most of the pollutants are being initiated in the combustor, developments of this component have been crucia l for emission abatements. In addition, with the progressive change in the fuel sources landscape, burners should now meet the challenge of the adaptability to promising alternative fuels along with a decreasing pollutant generation. On future gas turbine combustors, different objectives should converge. The emissions abatement is of primary importance for both power generation and aircraft propulsion systems. Single digit nitric oxides and carbon monoxide emissions are becoming a reality in combustors, in cluding new flameless configurations. Another motivation comes from scaling down the combustion systems. Having the highest heat release in the smallest volume is a priority in aircraft propulsion systems. This would reduce the weight, and so the overall f uel consumption. Reduced fuel consumption could also go along with improved efficiency requirements. In addition to these conventional trends, changes in the way we consider the fuel sources bring new considerations. Even though stability in the combustion system has always been a crucial point on aircraft combustors in which a flame should be anchored over a wide range of conditions during the flight, the localization and diversity of the alternative fuel


14 source candidates in ground power generation also c alls for a need in developing stable fuel -flexible combustion systems. Emission abatement has been mainly done in the last decade through the d ry l ow NOx technology [1] Different mechanisms have been identified in the generation of Nitric oxides such as the Zeldovich or t hermal NOx mechanism and the Fenimore or prompt NOx mechanism. Overall, NOx emissions are strongly linked to the flame temperature. Therefore the main concept would be to operate in a lean regime so that the excess of air will lower the f lame temperature. In addition, the mechanism of soot formation is very dependent on the local fuel to air ratio in the flame. Large gradients in the composition of the fuel air -mixture will lead to the formation of soot precursors. By mixing the air and the prevaporized liquid fuel upstream of the injector, a premixed combustion system delivers a nearly homogenous mixture decreasing significantly the amount of soot formed. However premixing flames offers a narrow range of stability compared to diffused bur ners, which makes lean premixed technology difficult to use in aircraft propulsion system. In term s of emissions, significant decreases have been observed in d ry l ow NOx system s burning n atural g as presenting a mole fraction of CH4 varying from 70 to 90%. The current energy context offers a promoting force for investigation of alternative source s of fuels such as biomass and co al. However coal -derived synthesis gas, mainly syngas composed of hydrogen and carbon monoxide present combustion characteristics quite different from the well -investigated methane. Moreover, fuel supply may become more and more local as the diversification of sources become a reality and therefore the composition of the syngas is expected to be highly variable. The


15 premixing concep t used in dry low NOx systems is particularly sensitive to the composition of the fuel used Currently; their performance is restrained to methane and their adaptability to the future fuel changes still needs to be assessed Changes in fuel composition not only affect the emissions and heat release performance, but the other components of the air path such as the compressor and the turbine. Increase in air flow rate due to a low er heating value of the burning mixture may affect the compressor performance. A lso, corrosive products could also be a major issue in material wear. Concerning the combustion itself, the oxidation chemis try plays a very important role, especially in premixed system s Operability and emissions performance are therefore greatly influen ced by the fuel composition. In premixing systems the mixture of air and fuel is flammable far upstream of the injection point meaning that if the ignition delay is not sufficiently controlled, ignition could occur in an unexpected section of the system a nd could have some critical effects. An even m o re critical phenomenon would be a flashback on the air supply lines. Finally, i f the chemistry dynamics are not sufficiently investigated, combustion limits could suddenly be reached (i.e. flameout) within sma ll changes in operating conditions A lot of efforts are now being done in order to investigate the oxidation of these exotic fuels through different experimental set ups in parallel with computer simulations. Information on the oxidation and pollutant for mation kinetics could be extracted from different measurements, such as ignition delay, laminar flame speed, and flame blowout. Shock tube experiments have been studied extensively and ignition delay correlations have been extracted for several fuels over a wide range of conditions using these systems. The ignition delay is also greatly influenced by fuel composition. These


16 experimental times could be compared with computer simulated values in order to develop a mechanism or a reactor model. Practically, a uto-ignition at an undesired place could also damage the combustor. Kalitan and co-workers have studied ignition delays of CO/H2/air mixtures behind reflected shockwaves at atmospheric con ditions and elevated pressures [2] They compared their experimental results to current kinetic models. This step is a critical precursor of the use of synthesis gas in power generation plant. Indeed, H2 and CO have particularly low flammability limits and the combustion rates of a synthesis gas mixture are strongly related to the H2/CO ratio. Shock tubes are actually based on a transient process but ignition delays can also be obtained from continuous flow devices and rapid compression machines. The laminar flame speed is a concept based on premixed laminar flame combustio n. It contains information on the kinetics of the combustion of the mixture but also on the diffusivity of its species and related thermo chemistry [3] Natarajan and coworkers burned CH4/H2/CO mixtures in Bunsen burner and wall stagnation flame reactors They performed f lame speed measurements with a charge coupled device (CDD) camera on the Bunsen flame, and laser Doppler velocimetry on the stagnation reactor. Premixed mixtures went from 5-95% for H2 and CO and up to 40% for CH4, and measurements have b een done on preheated mixtures up to 700K, and at pressures ranging from 1 to 5 bars. Measurements could then be compared with some detailed methane oxidation mechanism such as the Gas Research Institute mechanism version 3 (GRI Mech 3.0). Detailed mechani sm provides with kinetic rates of elementary reactions such as thermo chemistry of the species involved. The GRI 3.0 describes a


17 detailed mechanism appropriate to gas natural combustion. It includes the NOx formation and reduction but does not describe soot formation [5]. In addition, lean blowout measurements are often done on premixed pre-vaporized swirling flame combustor devices. Based on the concept of chemical time limited reaction, the residence time of the reaction at blowout could be linked to the reaction rates and compared with a reactor network model. Zhang et al. studied the effect of the composition of H2/CO/CH2 mixtures on lean blowout in a premixed gas turbine combustor (Zhang et al. 2007). However in this type of device, the lean blowout theory holds as long as the reaction time is larger than the mixing time scale. A non homogeneous reaction zone will yield to other combustion regimes such as flamelets into eddy. In order to uncouple the convective and diffusive physical processes from the c hemical reactions, a very strong mixing should be achieved such as in a well stirred reactor (WS R). The w ell -stirred reactor theory has been ex tensively used as a modeling tool to investigate reactor behavior. The first experimental device considered as ap proaching the well -stirred reactor concept has been designed by Longwell and Weiss in 1953. Their primary motivation was to extract global reaction rates over different air/hydrocarbon mixtures [6] In a well -stirred reactor, the composition in side the rea ctor is homogeneous. Therefore the exhaust composition is identical to the composition inside the reactor. The idea is that a residence time of the gases inside the reactor could be extracted which would put together the influence of the incoming flow, the temperature, the pressure inside the reactor and the volume of the reactor. This residence time could be given by the expression below:


18 .resVPV t mmRT (0 1 ) This time would represent in theory the time allowed for the chemical reaction to progress toward completeness. Just before blowout the maximum reaction rate is obtained. The w ell -s t irred reactor developed by Longwell and Weiss was brick insulated and a premixed pre vaporized air -fuel mixture entered the reaction zone through central orifices creating sonic jets. The mixing forces created by these high velocity jets provided a strong homogeneity. Blowout measurements were performed for rich and lean mixtures. Later, some other well stirred reactors have been designed such as the hemispherical stirred reactor developed by Wright [7] and then Blazowski [8] or some conical reactors [9 -1 0] A significant advance has been made by the development of the toroidal jet stirred reactor developed by Nenniger [11] In his device, a plug flow reactor was used in line with the WSR. An investigation of the mixedness of the reactor has been carried o ver this device and reveal ed some very small mixing times of 2 104 s for a reactor residence time of 5ms. Nenniger ran f uel rich air/acetylene and air/toluene mixtures in his 250mL reactor. Soot, PAH, light hydrocarbons and oxygen were probed over the length of the plug flow reactor corresponding to different residence times. This reactor was then extensively used by different researchers, for examples, to study the stability and emission of gaseous fuel [12] the emission characteristics of different liqu id fuel s [13] and later to characterize soot inception on ethylene/air mixtures [2004] Well stirred reactors have been a useful tool to understand emissions and soot formation. They represent an advanced combustion system that is primarily limited to kinetic


19 investigation purposes. The brick material and the shape of the combustor are not yet suitable to practical system for power generation. Moreover, the sonic stirring is not possible on high power system with considerable flow due to important pressure drops and material buckling limits. The work presented in this thesis aims at developing a system which is inspired by the well stirred combustor systems but which may also be applicable to actual combustion systems, a micro turbine system used for semi cycle experiments. This system would need to be highly fuel flexible so to obtain combustion data over a wide range of mixtures and conditions and accurate control should be possible over a range of equivalence ratios, the air flow rates, and the temperatures. Also the combustor flow path should be simple enough to enable combustion modeling by reactor neural network so that reaction mechanisms can be investigated. And of particular importance, a state of art diagnostic should be implemented for combustion products characterization.


20 CHAPTER 2 STATEMENT OF SCOPE Objectives of the combustion Rig Reaction limited operation. In liquid fuel combustion, three consecutive processes have to be considered. The breakout and evaporation of the liquid fuel into small droplets comes first, then the mixing between the fuel and air over the combustion volume and finally the oxidation reaction. In most of the practical systems, one of these three processes limits the consumption of the fuel. The coupling between these processes makes the chemical mechanism very difficult to investigate in a traditional non -premixed flame where composition varies thr oughout the flame. Indeed in order to estimate reaction rate parameters, the concentration of the reactant and the temperature have to be known. The collision theory and empirical results provide the reaction rate expression for a bimolecular elementary r eaction step: ABCD (0 2 ) i n which the rate of reaction is: [] [][], dA kAB dt (0 3 ) and k is expressed by: ()exp(/),b akTATERT (0 4 ) where A and b are some constants and Ea is the activation energy. If the equivalence ratio, flow of the reactant and temperature varies significantly in the flame, the access to kinetic rates would be very difficult. The ideal for fundamental


21 understanding would be to have an homogeneous mixture of vaporized fuel, air, diluents and products at a controlled temperature all over the reactor volume, so that one can compare the product composition (emissions) to a zero-dimensional set of inputs (temperature, flow of air, equivalence ratio, dilution ratio, pressure) for a given fuel. The values obtained could then be compared to computer simulation results based on a chemical mechanism and the well -stirred reactor theory. The reaction should be reaction rate limited. An ideal combustor would have its mixture instantaneously vaporized and mixed as soon as it enters the reactor volume. This objective has been approached by preheating the fuel and then spraying it into a vaporization chamber where it mixes with the air at an elevated temperature. Within a sufficient time between the spray and the injection, the liquid fuel should vaporize completely. Then the mixing into the combustor chamber is achieved in Jet Stirred reactor through a significant number of small orifices creating some sonic jets, stirring the reactor vigorously. In a gas turbine system, the sonic condition is hardly achievable. In fact the negative pressure gradients on the compressor liner wall due to a negative relative pressure would lead to wall buckling. In fact, the liner is generally very thin for weight reasons and the buckling limit is reached before the son ic regime. Stability The stability could be defined as the ability of a system to maintain combustion over a certain range of conditions. The stability performance can be considered either as the range of fuel air ratio over which a combustion regime coul d be sustained in the system or either as the maximum air rate the system could accept without extinction at a given composition This characteristic of the burner would be an objective for fuel flexibility. For example, the wider the operation range, the better the oxidation chemistry


22 could be understood. The phenomenon of blowout in which the flame is no longer attached to the injector and is b lown out of the burner occurs at the limit of the often c alled static stability range It is opposed to dynam ic stability or instabilities describing pressure oscillations which created damages in the combustor materials due to heat release rate variations [15] In an ideal well -stirred reactor, the stability loop would be the widest for a given premixed mixture compared to all practical systems. Longwell and Weiss have published stability loops for a variety of fuels usin g their spherical reactor [16] There are two dimensions in a stability loop. To understand the static stability concept, the energy balance in a well stirred reactor is illustrated on F igure 2 1 in which the curve trends are extracted from Nennigers PhD dissertation [11] The heat generation is a complex function of the reactor temperature which depends on the complete reaction mechanism. The heat losses curve is a straight line if we consider an adiabatic reactor and a constant heat capacity mixture To solve the energy equation in steady state means to equalize the heat generation and the heat losses so that the solution of the system is giv en by the intersection of the two curves. The lowest point corresponds to the case of a non -reactive mixture and the highest correspond to a flame regime. The intermediate point does not provide a stable solution as it is located on the decreasing rate of the heat generation so that steady state could not occur.


23 Figure 21 Heat balance in a Well Stirred reactor The first dimension is the equivalence ratio. For a given mass flow rate, combustion could occur only between two equivalence ratio limits. On the lean side, the dilution of the air/fuel mixture with some additional air will decrease the reactor temperature, having a negative effect on the reaction rates. Indeed, the dilution effect will increase the slope of the heat loss curve and the higher i ntersection point will decrease on temperature. It can be seen on F igure 1 2 that in theory, there will be a point in which a further rotation of the heat generation curve will prevent the crossing of the two curves and so the only solution for the reactor will be at the feed temperature. On the rich side, blowout is still under active investigation, but the formation of soot and polycyclic aromatic compounds enhance the heat transfer to the wall due to blackbody radiation of the burning particles and the heat losses are getting significant However, rich blowouts are more complex due the water -shift gas equilibrium and the fast reaction rate of H2, the main product of the breakout of a rich hydrocarbon air mixture.


24 Another dimension for static stability is the one associated with the mass flow rate. For a given equivalence ratio, if one increases the air flow rate, the time allowable for reaction or residence time will decrease. The reaction will not be complete, thereby decreasing the reactor temperature, but in the same time the concentration in reactant will be more important. The concentration effect will outweigh the temperature drop, and the overall reaction rate will be faster and the flame will sustain. If one keeps increasing the flow, there will be a point where the temperature decrease effect on the reaction rate will be more important than the reactant concentration increase, and combustion will no longer occur. Computer -based kinetic simulation can also be used in parallel with blowout data. Comp uter simulation will provide the steady state energy equation solutions over a range of flow rate and equivalence ratio. If one is able to detect blowout with precision, one is provided with a valuable data set. Blowout is highly dependent on reactor press ure. The larger the pressure, the larger the air flow will be at blowout. Indeed the reaction rates are directly related to the concentration of the reactant given by the pressure and the mass fraction. The easiest way to get blowout data would be to maint ain a low pressure in the combustor volume so that blowout occurs at the minimum flow rate without being limited by the compressor or other devices. Moreover, by running close to the blowout point means also running at lower temperature. These non equilibr ium temperatures are also called kinetic temperatures. Therefore a direct access to the flame temperature by means of insulated thermocouple is simplified. On a practical system, the stability loop is narrower. This is actually due to the heat losses on the one hand, and to the incomplete back mixing between the incoming


25 flow and the combustion product on the other hand. These two phenomena prevent the fresh mixture from reaching a sufficient temperature, which would for a given residence time lower the com pletion of the reaction. Inside the flammability limits a mixture will blow out if not enough heat is transferred to raise the incoming mixture to the auto ignition temperature. Different techniques have been developed in order to stabilize the flame in a practical gas turbine combustor evolving in a premixed mode Bluff body flame holders and swirling vanes are used in ground based burner s In fact, a low velocity region would help the flame to sustain, as it would transfer heat to the fresh mixture by mea ns of the recirculation of the hot combustion gases. In the aircraft industry, sequential burners are used as a way to provide better stability. A first burner is fuel rich and the incomplete combustion products go into a second burner provided with fresh air to terminate combustion. In addition s taged combustion in which two or more mixture s of different equivalence ratios are injected at different positions, has shown significant stability improvements. In diffusion flames where eddies of fuel burn surro unded by air, the overall lean operation hides local rich-fuel regions which stabilize the flame. However, this type of combustion produces more pollutants and the reaction is vaporization -rate limited. Actually the stability and stirring criteria are one, hence a good mixing and a control over the heat losses is very important in the design. Control and diagnostics As discussed in the introduction, there are different ways of obtaining the kinetic parameters involved in the oxidation mechanism using some experimental devices. For all of them the accuracies on both the control over the inlet conditions and the measured variables (product composition, ignition delay) are critical to extract valuable data. The supply lines of air and fuel should be independen tly metered for mass flow


26 and temperature. As the equivalence ratio played a very important role in the oxidation chemistry, only small errors can be tolerated. Also, an access to the reactor temperature would be very valuable. It would allow calculating the global residence time defined in E quation 1-1. However, an ideal wellstirred reactor could reach a very elevated temperature which is the adiabatic flame temperature of the mixture if the reaction evolves toward completion. Indeed adiabatic flame temp eratures for air -fuel mixtures can reach more than 2000K at stoichiometric condition, with a peak for acetylene -air mixture at 2773K. Techniques for flame measurements include direct thermocouple measurement with a sheath insulated probe of platinum/ r hodium which could be used up to 2600K. It also includes laser -based techniques such as coherent anti -s tokes Raman spectroscopy (CAR S). CARS involve the interaction between the laser wave and the vibrational spectrum of a molecule in a coherent manner and a dir ected signal could be collected. The difficulty of this technique stems from the fact that the pump and probe wave (often the same signal) should be finely tuned to be resonant with a single vibrational transition of the sample. In addition, the classical concept of WS R implies an adiabatic reactor. Even though this condition is approached well using brick insulation, this material is not acceptable in a practical system. Indeed fusing and crumbling of the liner would damage the turbine blades. In the Longw ell s study of hig h temperature combustion rates [6] the liner had to be changed every 10 runs at very rich and very lean conditions, and every 2 runs for near stoichiometric equivalence ratio. As nickel -based alloy cannot stand more than 1400K, active cooling would be required in a system adaptable to a real gas turbine


27 system. Since heat losses could be implemented in a computer well -stirred reactor model, heat losses could be tolerated as long as they are precisely evaluated. In a well -stirred reactor, composition at the outlet is supposed to be identical to the composition inside the reactor. Based on the inlet flows composition and temperature on the one hand, and on the reaction mechanism and kinetic rates in the other hand, composition within the r eactor volume could be evaluated using a computer -based model. In order to understand oxidation chemistry, an iterative process could be done in order to match the computer simulation to experimental results under different conditions. Therefore, diagnosti cs over the product composition is crucial. Also the product composition should not change considerably between the reactor volume and the position of the probe or this change should be controlled. Nennigers toroidal jet stirred reactor includes two sections, a well stirred r eactor and a plug -flow reactor [11] Indeed, the well -stirred reactor was followed by a tube in which combustion kept evolving toward completion. In that case different sections over the length of the P.F.R will correspond to different residence times. A sequence of W.S.R and P.F.R could be simulated as well as a sequence of several WS R. Moreover, species composition diagnostics could be of different natures depending on the processes one wants to understand. These are generally of two types, real-time or gas sampling. Gas samples could be extracted from the reactor and composition analyzed by different methods such as gas chromatograph -mass spectrometer (GC MS). The development of laser based diagnostics has provided different real time techniques, especially on soot diagnostics such as Laser Induced Incandescence (LII) which can provide with particle size distribution and soot volume fraction [17]


28 Fuel and gases The development of a reactor for kinetic investigation of exotic fuels i ncludes a certain number of steps. Before running this fuel into the new system, the reactor characteristics themselves have to be investigated. Apart from the aerodynamics of the combustor, often studied for practical systems with planar laser induced vel ocimetry or using cold flow techniques, the kinetic model created for the combustor has to be adjusted on the basis of a reference fuel. Detailed oxidation mechanisms of several fuels have been developed at Lawrence Livermore National Laboratory (LLNL) or by the Gas Research Institute (GRI). The current interest in liquid fuel oxidation and the publication of two versions of a detailed mechanism for n heptane by Lawrence Livermore laboratory led our choice for n -heptane as reference fuel. Indeed, Curran and coworkers at LLNL developed this mechanism based on rapid compression machine experiments, reflected shock wave ignition delay in shock tube, and well-stirred reactor experimental results [18] The range covered by this mechanism goes from 1 to 42 atmosp heres, temperature from 550 to 1700 K, and equivalence ratio from 0.3 to 1.5 and the nitrogen -argon level of dilution from 75% to 99%. Recently, Herzler and co-workers listed different experimental studies performed on nheptane [19] However most of the e xperimental studies presented here have been done relevant to high charge compression ignition engines (HCCI) and diesel engine by their range of pressure and temperature. A second step of the implementation would concern some gaseous fuel. As for liquid fuel, the reference fuel, methane has been chosen. The GRI 3.0 Mechanism developed originally by the Gas Research Institute, provides a detailed mechanism for methane [4] In addition LLNL provides two mechanisms on straight chain hydrocarbons


29 to respectiv ely model NOx production with the first one, and soot and polycyclic aromatic hydrocarbons formation with the se cond one [20 21] A third step would be the test of exotic fuels and the investigation of different levels of dilution. Liquid fuel testing could include some synthetic fuel, for example from a Fisher -Tropsch (F -T) Process. By this process, synthesis gas obtained from biomass or coal gasification and composed mainly of CO and H2 are transformed into mainly straight -chain hydrocarbon. For selecti vity reasons, F -T synthesis takes place at low temperatures (150300 C) and catalyst su ch as cobalt and Iron are used [22] This process was invented in the 1920s by the two researcher s of the Max Planck Institute who gave their names to it. At that time, Germany was very poor in petroleum product compared to other foreign nations (U.S.A, U.R.S.S, United Kingdom), but very rich in coal and lignite. Then with the militarization of the third Reich in the 1930s, a lot of effort was put in order to process li quid fuels for the military. Therefore, production of F T fuel culminated during the war before the bombing of a large part of German oil factories. The United States showed in recent years a great interest in this process as part of the CO2 capture technology. The Energy and Gas Dynamic System Laboratory at UF tested some F T synthetic fuels and their operability has been demonstrated on a micro -gas turbine system [23] This process is closely linked to the gasification process upstream and raises a huge s cientific interest. Indeed, in the scope of reducing the CO2 emission and so the human signature on global warming, the development of solar powered plant, and the investigation of catalysts such as Iron could increase significantly the energy content of s yngas using the water -shift equilibrium. Therefore, an interest in syngas oxidation chemistry is raised and has been investigated through different studies


30 [3 5, 15 ]. For these reasons above, runs with mixtures of CO/H2/CH4 and air representing different qualities of syngas is planned and especially low H2 content mixture with a CO/H2 ratio greater than 5. In the current low emission research, the concept of vitiated combustion has been investigated for its NOx and soot emission reduction. In fact, a high level of dilution, in parallel with a high air inlet temperature creates a combustion regime characterized by a non -luminous flame, a homogeneous temperature, and a distributed reaction in the pr imary zone [2429] The inlet conditions could be in theory achieved with a recirculation of the hot exhaust gas. However in most system, the cooling of these exhaust gases has a detrimental effect on cycle efficiency. W.E Lear developed an innovative micro-turbine cycle where the cooling of the exhaust gases is pr ovided by an absorption refrigeration plant [ 30 ] This semi -closed cycle micro turbine has been derived to give the power water extraction and refrigeration system (PoWER) offering the advantages of a simultaneous generation of electricity, water steam, liquid water and refrigeration in parallel with lower emissions. On this system, a sharp decline in luminosity of the flame has been observed with the transition to t he flameless combustion regime [31] However, the studies carried on soot emission were cond ucted with commercial low -resolution optical equipment. The system designed in this thesis could be used for the development of state of art diagnostics technique on soot. Indeed the laboratory scale and the instrumentation would give a more flexible contr ol over the combustion conditions, and the optical access at the exit of the primary zone would offer an interesting position for soot characterization. Also, in the gas turbine combustor system, a lot of physical processes are coupled such as vaporization of the fuel, mixing


31 and reaction. If one is able to uncouple these processes, one could be able to investigate the fundamental effect of dilution on the combustion regime. Therefore, the dilution of the flammable mixture with gases will be implemented. The products of hydrocarbon combustion at equilibrium are mainly composed of nitrogen, carbon dioxide, water vapor, carbon monoxide (for very lean or rich mixture), oxygen (lean mixture), and hydrogen (rich mixture) due to the water -gas shift and nitric oxide. Using a CANTERA re actor model developed by Singh [32] a reproduction in an open system of an exhaust gas recirculation could be achieved setting the inlet composition close to the simulated results. As a summary, the fuel and gases that will be implem ented in this burner are: Air, as the oxidizer N -heptane (C7H16),as a reference liquid fuel Natural Gas (mainly CH4), as a reference gaseous fuel and a synthesis gas component Carbon Monoxide (C O), as a synthesis gas component and a diluent Hydrogen, as a synthesis gas component and a reference fuel Carbon Dioxide (CO2), as a diluent Water vapor (H2O) as a diluent Nitrogen (N2) as a diluent and liquid carrier for water vaporization From the considerations above, one can define the guidelines of the combust ion rig. The combustor geometry and the injection system has to provide a fair amount of mixing giving a good homogeneity of the mixture inside the combustor zone under investigation. Comparison between the combustion characteristics of the actual system o n a given fuel and the results from a widely investigated kinetic model representing this


32 combustor should be possible through adequate diagnostics techniques. Finally, the i nvestigation of the combustion of the different fuels and gases listed before shou ld be possible over the widest possible range of controlled inlet conditions (temperature, composition, flow rates) Constraints and Limitations The burner to be implemented is this study is integrated into a bigger system involving the fuel delivery, the air delivery, and the exhaust. Some technical limitations due to the other devices, which are part of the entire combustion system, will have to be considered in the design. The important parameter that must be determined for the design will be the reactor volume. In fact, given a volume of reaction and a range of residence time of interest, one can calculate the air flow needed for combustion and the quantity of liquid fuel needed. Flow rates The current laboratory cannot power an electrical device which requires more than 7.5 horse power. The compressor and the air heaters are the electricity consumers. A compressor of 7.5hp could in general deliver 24.2 ACFM (Actual Cubic Feet per Minute) at 175psi corresponding to 284 SCFM (Standard cubic feet per minut e) using standard conditions at 60F and 1 atmosphere. A dryer could be used to get rid of humidity .The SCFM units are in fact a mass flow rate as one can convert a volume into a mass given the temperature and the pressure so that 284 SCFM would be around 130g/s using the equation V std act ACFMSCFMactsat stdPT V PPT (0 5 )


33 where is t he relative humidity and Pact, Psat, Pact respectively the actual, saturation and standards pressures. As discussed before, one will also need some fresh air for cooling the hot material of the combustor. A heat transfer model must be programmed in order to estimate this flow at a given volume. The flow rate may be also limited by the fuel consumption. In fact by safety reason, one cannot store more than a certain volume of fuel. For this reason, the reference fuel consumption in continuous operation is li mited to 1 gallon per hour. Pressure In combustion, the pressure has a very important role on the oxidation reaction rates. In fact the probability of collision between two molecules will increase considerably with a shorter mean free path due to a pressu re rise and so a rise in number density. However the effect on an elementary step will depend on its molecularity defined as the number of colliding molecular entities In term s of blowout and stability, if one decreases the pressure on a given system th ere will be a limit where the flame will blo w out. By setting the pressure at a low level, one will need less air flow to reach an incomplete reaction regime and then a blowout. However, as the pressure decreases the flow required gets smaller and smaller and so the bulk effect of the heat losses become s very important. If one recalls the energy equation in steady, one can get the expression: ,p lossesT mCq x (0 6 ) where lossesq is a linear heat flux. This latter term includes a constant term with respect to the mass flow of air representing the radiation heat flux and a power function of the


34 mass flow rate through the convective effects. The convective heat transfer rates in a gas turbine combustor are described in more detail in the heat transfer section of the C hapter 3. By decreasing the mass flow rate, the temperature gradient will get larger according to E quation 2. 5 since there is a fixed part of the heat losses and a variable part in which the dependence on the mass flow rate is elevated to a power less than one. Moreover a negative relative pressure would raise som e technical difficulties. Indeed, in this case, we would need to use a vacuum pump downstream of the combustion system to raise the pressure to the ambient and so drive the flow to the exhaust. This type of system is shown in F igure 2 2 A) with an arbitrar y atmosphere pressure in the combustion primary zone. The exhaust gases being very hot, the control over the exhaust will need some expensive temperature resistant devices with a large volume flow rate rating. For these reasons the design will try to get the smallest but positive relative pressure as possible in the combustion chamber. In this case, the system will be similar to the one in F igure 2 2 B) A compressor will alternatively drive the flow from the ambient to a tank on the basis of a duty cycl e. The pressurized tank at a pressure up to 180 psia will deliver the air to the injection system at a decreased pressure using an expansion valve. The c ombustion pressure will be over the ambient so that the exhaust flow could be driven to the outside. No w, the combustion pressure will be dependent of the volumetric air flow and the pressure drop coefficient induced by the downstream exhaust pipe. It could even be controlled through a valve. Some advantages in this system would be the possibility to use th e pressurized tank receiver to provide air to a small machine shop. The laboratory in which the combustion rig will


35 be installed is new and without compressor so the choice of the compressor and the design of the air supply is also part of the work realize d in this thesis. A B Figure 2 2 Absolute pressure evolutions (in kPa) along the flow path of a combustion system burning at negative pressure A) and a positive pressure B ). Volume On the large size, the volume will be limited by the maximum flow deli vered by the compressor. The minimum volume limit comes from the heat exchange are -to volume ratio. As the diameter decreases, this ratio increases for an annular reactor. Moreover there is a point where machining gets very difficult as the combustor is m iniaturized. A kinetic simulation is presented on the following chapter in order to set the maximum volume size. In addition, a heat transfer model developed in C hapter 3 will estimate the heat losses on the combustion gas bulk temperature and determine if whether or not a volume is acceptable. Inlet and reactor temperature The inlet temperature is an important parameter for the stability of the reactor as well as the flame temperature or reactor temperature in a well -stirred system. Indeed, in


36 a WS R, the composition is homogeneous and so is the temperature. This is an idealized concept; however, in reality it can depend on heat transfer calculation s to predict the limitation in terms of wall cooling and heat losses. Indeed the higher the flame temperature, the higher the heat transfer rates and the better the cooling should be to maintain the material of the combustor at a reasonable temperature. As an ideal normal stainless steel alloy (SS 30 4 ) could in general hold a continuous 900 K, a nickel based alloy such as Inconel or Hastelloy could stand up to 1 100 K, and Molybdenum has a melting point at 2880 K. Considering the benefits of higher temperature, the hotter the inlet mixture, the more stable will be the flame. However for an air/fuel mixture in an idea l W.S.R, the increase in inlet temperature would imply an increase in flame temperature. In theory, within a sufficient residence time, a well-stirred reactor should reach the adiabatic flame temperature. Considering the reaction as adiabatic and that chem ical equilibrium is reached, the entire heat release should raise the mixture to the adiabatic flame temperature. In F igure 2 3 the adiabatic flame temperature is plotted at constant pressure for CH4/O2/N2 mixture for different mixture s and inlet temperat ures


37 Figure 23 Adiabatic flame temperature for mixtures of CH4/O2/N2 Figure 24 Composition at equilibrium for different methane air mixtures at initial temperature of 400K and a constant pressure of 1 atmosphere. These calculations were d one using Cantera, on a GRI3.0 mechanism by fixing the enthalpy and the pressure. For a given oxygen to nitrogen ratio, the adiabatic flame temperature is maxim um at a near stoichiometric mixture (O2/CH4=2). Indeed, the maxim um heat of reaction per unit mass is obtained as almost all of the reactants are consumed to give CO2 and water. The refore the flame temperature versus equivalence


38 ratio curve is a bell whose top is near the stoichiometr ic point The composition of the equilibrium mixture is shown on F i gure 2 -4 One can observe that for a stoichiometric mixture the products are essentially CO2 and H2O and N2, and so the heat of reaction is maxim ized On the lean side, one sees that a good part of the oxygen remains in the products, and so th e inert oxygen consume s some of the reaction enthalpy to reach the adiabatic flame temperature. As one go es toward rich mixtures, the concentration of CO which enthalpy of formation is lower than CO2 becomes predominant. For this reason t he combustion could be conside red as incomplete as not all the CO is completely oxidized in CO2 following the reaction: 2222. COOCO (0 7 ) T he water -gas shift equilibrium is also involved on rich combustion leading to the formation of H2 in the products. The reaction involved is: 222COHOCOH (0 8 ) Even though the mass fraction is low, the lightness of hydrogen hides a large molar fraction of 17.2% on F igure 2 4 .As one goes to ward a pure oxygen/fuel mixture, there is little or no inert species and so less heat is need ed to raise them to the final temperature, and therefore the remaining reacting species are heated to a higher temperature. On the other side, as level of dilution by inert gases such as nitrogen is increased the adiabatic temperature drops directly linked to the thermal effect of dilution. This type of combustion is of particular interest since the reactor is at a reasonable temperature and the gradient between the inlet temperature and the adiabatic flame temperature is


39 fairly small promoting a homogeneous reaction zone. This is one of the ideas behind f lameless c ombustion or m oderate and i ntensive l ow oxygen d iluted combustion (MILD) as it is named in Italy. However the kinetic rates are very slow at 400K in a diluted regime which could not be observed from adiabatic temperature results. Therefore, the f lameless combustion concept is also called h igh t emperature air combustion (HTAC) in Japan. The objective is to have a high inlet te mperature together with a small temperature rise due to combustion. This high inlet temperature is typically obtained by an e xhaust gas recirculation as mentioned above and the hot exhaust gases are cooled down to 700-800K mix with the fresh air and are in jected back into the combustion chamber. Most studies performed on WS R based their measurements on an air/fuel mixture inlet temperature of about 400K. For a premixed system, the advantages of this 400K is that this is below the auto ignition temperature of most of the fuels which prevents ignition in the premixed section, and over the boiling point of the fuel, ensuring a prevaporization of this latter. Indeed the fuel in injected in the air stream and then the two mixes before being injected. If the a ir is at a temperature higher than the boiling point, heat will be transfer to the fuel to start its vaporization in the mixing section. Moreover, we want to prevent ignition upstream of the combustion chamber which would have damaging effects on the mater ials, so we do not want the air to reach the auto ignition temperature of the fuel. We will target a 400K temperature for the air over the entire range of flow rate achievable. Concerning the vitiated combustion regime (dilution with N2 and other gases ), w e want to raise the inlet temperature to 700K but we will limit the


40 total flow rate so that no additional heating power will be required compared with the fuel/air combustion regime. In Table 2 1 some important physical properties are listed for the diff erent fuel planned to be implemented in the system. These results are extracted from International Chemistry Safety Cards (ISCS) and are more oriented on safety than on fundamental combustion properties. However, they could indirectly provide information a pplicable to premixed combustion systems design. For liquid fuel, the flashpoint is defined as the temperature at which the vapor pressure of the liquid fuel will provide in the gas -phase a concentration of fuel molecules high enough so that it forms a flammable mixture with the surrounding air. Practically, it would be the minimum temperature at which one can ignite a liquid fuel under ambient pressure using an ignition source. When the ignition source is removed combustion would stop. Another temperature, important for both gaseous and liquid fuel is the auto ignition temperature. It is defined as the minimum temperature at which a flammable of air and gas mixture ignites spontaneously under a standard pressure (1 atmosphere). One should keep in mind that these concepts are empirical and of practical sense more than being fundamental values. Theory will tell that under ideal conditions, a mixture of air and fuel will always react after an infinite amount of time. In addition, s afety standards organiz ations often relate to flammability limits. These tests are carried on vessels that contain a gas of a given composition. A spark plug is then used to ignite the mixture and if the mixture is inside the flammability limits the flame propagates within the vessel volume. The energy that is provided to the mixture for ignition by the plug is called ignition energy. This ignition energy is also function of the temperature and composition of the mixture.


41 Table 21 Some physical and chemical properties of the diff erent fuels involved (NA: Not applicable) Flammable fluid ISCS 0001 ISCS 0023 ISCS 0251 ISCS 0657 Hydrogen Carbon Monoxide Methane N heptane Boiling point 253C 191C 161C 98C Density (liquid phase/water) NA NA NA 0.68 Vapor pressure at 20C NA NA NA 4.6kPa Density (vapor phase/air) 0.07 0.97 0.6 3.46 Flashpoint NA NA NA 4C Flammability limits (in % vol) 4 76 12.5 74.2 5 15 1.1 6.7 Auto ignition temperature 500 571C 605C 537 285C


42 CHAPTER 3 DESIGN OF THE BURNER System approach As described in the previous chapter the design of this entirely new combustion rig in a new building space will include all the gases and fuel supply lines and the exhaust. The air will be compressed, metered, and heated. The liquid fuel will be pumped, metered, heated and sprayed into the combustion air stream. The mixture of fuel and combustion air will be injected in the combustor pr imary zone Some of the air will go directly to cooling passages and then will mix with the combustion products. After combustion, t hese two streams will be exhausted outside. The different systems to design are represented in F igure 3 -1 The first step i n the desi gn will be the combustor itself, because its performance defines all the other component s of the system. Figure 31 Combustion System approach for premixed liquid fuel/air mixture


43 Combustor Geometry As described in the objectives, the geometry should allow a sufficient amount of back mixing in order to stabilize the flow and to provide a better homogeneity. Premixing of air and fuel in combustion has been seen in the past decades as a significant way to c reate low CO and NOx emissions in gas turbine combustors. Significantly low gradients in Fuel/air fraction of the mixture entering the primary zone prevent the formation of hot spots responsible for the formation of thermal NOx. For the same reason, a significant decrease in polycyclic aromatic hydrocarbons (PAH) production is obtained. Studies on premixed laminar flame s showed that soot occurred only above a highly rich equivalence ratio [14] However, premixing also brought stability issues in homogen eous mixtures, particularly as they operate near lean blowout. Stabilization techniques, such as bluff body and swirlers have been seen as a way to achieve significantly stable operation at low equivalence ratio. Re circulating combustion gases transfer heat to the cool mixture, raising it to the auto ignition temperature. Longwell and then Ballal and Lefebvre have seen the recirculation zone as a homogeneous reactor [33 34] The stabilization performance depends on the time for the cold mixture to enter the shear layer limiting the recirc ulation zone, the rate of entrainment of fresh mixture into the recirculation zone, and the residence time of gases inside the recirc ulation zone. However, in swirl -stabilized flames, vortex precession and breakdown phenomenon would be an important source of combustion instabilities so that active boundary control and Helmholtz resonators are often employed in current d ry low NOx technologies [1]. The idea behind the current design is to provide back mixing based on jet impingement and flow split rather than large vortices structures.


44 The combustion zone geometry created in this project has been elaborated to provide this low velocity recirculation zone. This geometry is shown on F igure 3 2 and Figure 3 3. The annular combustor presented here is made of a h ollow cylinder liner with a wall on one side and the exhaust on the other side A smaller tube is mounted and the mixture is injected almost radially somewhere along the length of the liner. The fresh mixture will impinge at a fairly high velocity into the central tube so that part of it will be entrained toward the rear wall and the other part will go toward the exhaust where it will be later diluted with the liner cooling air. This design will provide a cross flow between the injection stream and the comb ustion product s coming from the back of the reactor and so t he he at transfer and mixing will be enhanced. A B Figure 32 Normal A) and axial B) cut of the burner


45 Figure 3 3 CA D view of the primary zone of the burner In this design two parameters were incorporated to address geometry influences. T he inner tube can be replaced, changing considerably th e jet impingement configuration as well as the combustor volume. In addition, the injectors are cylindrical and incorporate a slight 10 injection angle with respect to the radial direction so they can be rotated to act on the flow split as show in figure 34 Figure 34 Flexibility in the injector design


46 The injection angle is chosen as a compromise between a good tangential distribution of the f low and a jet impingement speed. The swirl number has been calculated as defined in the equation: 2 22,o i o ir z r r z rUUrdr S Urdr (0 9 ) where U and Uz are respectively the tangential and axial velocities, ri and ro the inner and out er radius of the annular space. Considering the cold gas injection velocity and the combustion product axial velocity, the sw irl number could be approximated at 10. However, the high velocity of the cold mixture jet imply a stirred reaction zone more than a swirling flow In this design, air and fuel are injected at four different points around the circumference and the jets a re directed toward the center tube. The number of injectors has been seen as the simple way to provide a good homogeneity and a good control over the reactor composition by opposition to a liner with holes. A future step in the implementation could include injection of air trough 8 additional radial injection ho les. This would provide an ortho radial ly stag ed combustion. Staging has been a central concept in the COSTAI R burner development for NOx reduc tion in gas turbine combustors [35] The pressure drop c ould be metered to ensure the same jet impulsion at the injection. This geometry has been inspired from a system designed at Florida Turbine Technology by Mark Harris for non premixed combustion [36] It revealed a very good stability performance. Once the idea of the design is defined, some main parameters have to be fixed. These are the ratio between the diameter of the liner and


47 the diameter of the center tube influencing the jet impingement, and the position of the injection circle with respect to the length of the combustor acting on the size of the primary zone and secondary zone. Kinetic modeling The objective in this part is to determine a volume of the reactor which would allow reach of the targeting operation points. Indeed, the goal is to be able to control the residence time as much as possible in order to reach the regimes of interest. A parameter has been used for many years in order to correlate mass flow rate, pressure, temperature volume and is named the loading parameter This parameter p rovides a partly non -dimensionalized variable which is used in gas turbine combustor design to describe the combustion efficiency. Indeed Greenhough and Lefebvre correlated defined in the equation : () 300 a T n cm PVe (0 10) w here P stands for the reactor pressure, Vc the reactor volume, T the inlet temperature [37] The value of n was first correlated by Longwell and Weiss equal to 1.8 and used by Lefebvre and Greenhough as 1.75 [6 37] In order to determine the proper flow rate or residence time of interest, it is useful to model the combustor behavior so that one can get estimation in terms of blowout limits, combustion efficiency, and emissions over the different inlet conditions for the reference fuel. As the combustion process involves aerodynamics, heat transfer and chemical kinet ics, the modeling might be very complex numerically. Turbulent flow modeling is already a challenge but reacting turbulent flow adds a lot more complexity.


48 Turbulence modeling has been investigated through different approaches such as the k Epsilon model, the Reynolds Stress Model (RSM), the large eddy simulation (LES), or Direct Numerical Simulation (DNS). Combustion modeling often uses these turbulence models along with some simplified reaction mechanism in order to have a reasonable computational time. I ndeed, detailed mechanism such as GRI 3.0 involves more than 200 species and 3000 reactions and the calculation time becomes enormous if it is implemented on each cell of a CFD model. Eaton and co-workers achieved a review of the different comprehensive co mbusti on models established [38] However, for emission and stability research, detailed mechanism are unavoidable to understand the pollutant formation and especially the chemistry of P.A.H. Modeling with a detailed mechanism often implies working with ideal reactors of different kinds. The continuously stirred tank reactor (CSTR) model is defined under the same assumptions than a well -stirred reactor which are a continuous flow of products and reactants (steady state), a uniform composition throughout the tank and an exhaust composition identical to the tank composition. In a CSTR, the residence time varies with the flow of reactants, and the reactor is considered zero dimensional. In a plug flow reactor (PFR) model, the flow is one-dimensional, steady, and frictionless as well as no mixing is allowed in the axial direction. In a PFR, the farther from the inlet the longer the residence time is. Even though most of the practical combustors are very far from these ideal models, some zones of a combustor could be more and less linked to one or the other of these models Modeling of a practical combustion system is often achieved through a reactor network approach. It consists in doing a first CFD simulation on a fine mesh with basic chemistry of the combustor. Then depending of the goal of the model, some zones of


49 similar properties are defined such as velocity, temperature and composition. Then thes e zones could be modeled as a CST R with defined a parameter, and the flow between these zones can be estimated fro m integration of the velocity extracted from CFD results. A network of ideal reactors is so constructed and detailed chemistry model could be computed on each of these reactors. More details on the reactor network approach for combustor modeling could be f o und in a paper from Falcitelli [39] In addition, Lebedev and co -workers modeled a gas turbine combustor for emission research with a reactor network approach [40]. For the current project, in a first step, a simple reactor network model is based on a ph ysical model more than on a CFD approach. Mark Harris has highly contributed in both programming a useful tool that provides an easy to use interface, and sharing his experience in running combustion rigs when setting a combustion model [36] Actually if o ne observes the reactor geometry, on F igure 32 and Figure 3 3, one can separate the combustion zone of the burner in three zones represented on F igure 3 -5 Indeed, the flow will impinge on the central tube, creating a flow split. Some of the flow will go toward the back wall and come back to mix with the fresh mixture. Then it will move downstream and start to be diluted with the cooling air. The primary reaction zone corresponds to the annular reaction volume located from the rear wall to the injection. T he secondary zone will correspond to the annular reaction volume located between the injection and the end of the liner. Finally, a short dilution zone is considered out of the liner and which shape is also annular. It is important to notice that in this C TSR network model, the dilution of the cooling air with the combustion products does not appear. Indeed, diagnostics were implemented to probe right at the exit of the liner. The cooling


5 0 passages have been designed to minimize the interaction and mixing be tween the cooling air and the combustion products close to this zone. There may be some reactions downstream as the combustion may be incomplete in the liner, and some of the intermediates products may keep reacting while mixing with the cooling air. Howev er the current focus is primarily on the reaction occurring in the annular reaction zone. A B Figure 35 Primary (red), secondary (green) and dilution zone (blue) us e d in the reactor network model A) and flow transfer between WSR according to the model 1 B). In the simulation process, the flow rate of air, the inlet temperature and the pressure are set to a certain value and a reactor total volume is also defined. A CSTR network simulation is performed for a given flow rate of fuel. After, the fuel fl ow rate is raised by a fixed amount and a new computation is performed. The process stops when a simulation results gives a reactor temperature equal to the inlet temperature. This means that the flame blew out. Two sets of computations were performed. The first one was done at a loading parameter of 1moles/s/liters/atm1.8 and a second one at two times


51 this value It should be noticed that for these runs, a total reactor volume had to be fixed as required by the simulation, and so was set to a 1L. However t his volume does influence the calculation only by its ratio with the mass flow rate. Therefore, one can extract the l oading parameter and use this one to evaluate which volume would produce the same condition (exit temperature, and composition) given the m ass flow rate available in the laboratory. Concerning the reactor model used, the flow distribution was set as described in F igure 3 -5 Again, the way the flow is split comes from a good correlation between this model and the experimental results ob served on a gas turbine combustor whose geometry was close [36] The term exhaust is not really appropriate as the combustion gases will be further diluted downstream of the liner with the cooling air. Figure 36 Temperature and residence time obtained at LP=1 (model 1)


52 The results of these simulations are plotted in the F igure 3 -6 and Figure 3 7 The equivalence ratio is based on the fuel -air ratio defined with the injection air only (excluding the cooling air ). The exit temperature was the temperature of the third CSTR. It is important to notice that this temperature will not be equal to the exhaust temperature as the combustion products are further diluted with the air from cooling passages of the liner and the tube. Finally the residence time in the prim ary zone is calculated using E quation 1 -1. As the flame blows out the reactor, the temperatures drops and therefore the residence time jumps due to the fact that the density of a cold mixture is low. In fact the mixture being no longer ignited, no signific ant reaction occurs and the only solution available will be the one of non-reacting flow with the exit temperature equal to the inlet temperature. In addition, the equilibrium temperature is plotted, which represents the temperature that would be obtained if an infinite residence time is allowed. This latter was computed using CANTERA commanded by Matlab. The gas properties were set as followed: the temperature was set at 367K as in the CSTR network, the pressure at one atmosphere, and the composition was varied over the equivalence ratio range of interest. The solution corresponds to equilibrium reached at constant enthalpy and pressure. The constant enthalpy condition accounts for the adiabatic condition of the reactor.


53 Figure 37 Temperature and re sidence time obtained at LP=2 (model 1) In the Figure 3 8 and Figure 3 9, one can notice that the range of global equivalence ratio goes from 0.25 to 1 for a LP equal to unit y, and from 0.25 to 0.8 for a L P of 2. As expected the stability limits get narrow er as one increases the load, and so a decrease in the residence time. However it is noted that this narrowing affects almost exclusively the richer side. In addition, the two kinetic or exit temperatures remain close between the two regimes at lean mixtur e, and then the gap between them grows steadily as the data moves toward richer mixtures. There are two important remarks that one should raise concerning these figures. The stability limits of the reactor for a LP of unity ranges from 0.35 to near sto i chi ometry. It was expected that these blowout equivalence ratios would be much larger on both the lean and rich side with similarity with a single CSTR model. This may be explained by the fact that these results were obtained with a model in which 100% of the fuel is going to the primary zone so that the


54 local equivalence ratio in the primary zone will be doubled compared with the global equivalence ratio. Hence, the rich blowout would be around 2 considering the local equivalence ratio for and a L.P of 1. In addition, the results do not have the bell shape for the exhaust temperature as one would expect if one thinks of the shape of the adiabatic flame temperature for n heptane. In the results, the temperature keeps increasing as the equivalence ratio goes up The effect of decreasing a short residence time is not the same for the two edges of the stability limits. This could be explained if one considers the breakdown of straight chain hydrocarbons into carbon monoxide and hydrogen. Indeed, a tool to model hy drocarbon combustion consists of breaking down the fuel into CO and H2, and then to consider the oxidation mechanism of the two fuels formed. The breakdown of fuel will often be considered as the limiting step, since its time scale is usually larger than t hose of CO and H2 oxidation reactions. The global fuel breakdown reaction could be modeled by: 222 2 xy y CHxOxCOH (0 11) In this equation, one could intuitively consider that as the hydrocarbon -to -oxygen ratio is increased, the breakdown will give more H2. The oxidation chemistry of hydrogen being faster than the one of CO, this could explain t he smaller effect that a decrease of residence time would have on reaction completeness.


55 Figure 38. Exit temperature and residence time in the primary reactor for two different loadings simulated using the model 1). Figure 39 Temperature evol ution in a WSR for a given composition Blowout


56 However, the sharp decline in temperature is not easily explained and should be investigated. Indeed, obtaining accurate results near the blowout limit is a delicate task since one has to go through a region in which 3 solutions exists at the same time, two of them are stable and one is unstable as shown on Figure 3 9 If one considers a given mixture of air and fuel, and starts at very low mass flow of air, the temperature will decrease as the mass flow rate is increased until the flame blows out. Starting on the other hand, if one considers an infinitely small residence time; the temperature of the reactor will be close to the inlet temperature. As the mass flow rate is increased, the temperature will increase very w eakly The mixture will stay fairly low even for mass flow rate larger than that at blowout on the forward approach. A minimum residence time is reached where the temperature jumps suddenly to the upper s ection. The portion of triple solution complicates t he approach of the blowout since this unstable solution may either jump into the hot side or the cold side. Indeed, the unstable solution being fairly close to the upper solution, a n inappropriate first guess could easily lead to an extinguished flame even if a real solution exits One may have to adopt a scheme in which the results from one simulation are used as a first guess for the next simulation. This approach has to be considered in the future when the experimental results will be compared to that of the numerical model. However, we will consider these preliminary results as satisfactory for our purpose here, which is to provide the exit temperature and some first stability limits. Indeed, a more sophisticated solving scheme may give wider stability limits. A third set of computations were performed at a higher loading, i.e. 4 moles.s1.liters-1 atm1.8. This time the scheme was changed to the one defined in F igure 31 1


57 where half of the fuel is injected in the primary reactor and half of the fuel is i n the secondary CSTR. In F igure 3-1 0 one can see that the results are quite different from the previous reactor network model. This time the stable regime goes from 0.5 to 1.7 and is more centered on the stoichiometric condition. The residence time goes d own to 0.3ms in the primary reactor. The exit temperature presents the same trend as in the other model, and is monotonically increasing with the equivalence ratio. The stoichiometric flame gives a temperature of about 1500K at stoichiometric condition and a temperature difference of 550K with the equilibrium temperature at an equivalence ratio of 1.3. However, these results raise the question of an exit temperature higher than the adiabatic flame temperature which needs to be further investigated to determ ine if this computationally -related or if the reaction is actually under a super -adiabatic regime when exiting the reactor. The current interest in having the possibility of running at low residence time is of two orders. First, the stability of this burner can be characterized Then if the reactor appears to be very stable, the strength of this geometry in stabilizing the flame will be demonstrated. Indeed, combustion of low heating value gases raises some static stability issues which make their use in gas turbine combustor very challenging For instance, running low H2 syngas (fraction of hydrogen less than 15%) is difficult due to the very low chemical time scale or auto ignition delay [15] Th is means that the mixture needs a large residence time in order to ignite and transfer the heat of combustion


58 Figure 310 Temperature and residence time obtained with computer simulations (model 2) Figure 31 1 Flow transfer between CST R s according to the model 2. The second motivation in reaching low ex it temperatures at stoichiometric conditions is in the fact that the reaction do es not reach equilibrium and some intermediate products exit the combustion zone of the reactor and can be probed. This would provide insight into oxidation chemistry of the f uel tested and data to retro feed a


59 reactor network model. At a large reactor residence time, the inefficiencies of the reaction are mainly due to the lack of mixing, to the heat losses and to the poor vaporization of the fuel as described in the W.S.R em issi ons study of Zelina and Ballal [12] In contrast at low reactor residence time, or higher loadings, the inefficiencies of the reaction will be also due to the slow chemistry part of the oxidation mechanism which reduces the completeness of the reaction. Moreover, the higher the flow, the lower the impact of the heat loss es and the better the mixing is. A third interest consists in running at lower temperature s, offering the possibility of direct probing of the flame temperature. As mentioned before, a sheath insulated probe of platinum/Rhodium could be used up to 2600K and with even further reduction in temperature, type K thermocouple s could even be used since they have a 1600K rating. In addition, low temperatures will make species probing easier us ing extractive gas sampling connected to g as c hromatograph-m ass s pectrometer systems or a commercial gas analyzer for hydrocarbons detection as a cheaper technology. Flame temperature together with combustion products composition will provide very import ant data in order to understand the reactor behavior. Comparison of the results with some values obtained using a neural reactor n etwork model may allow for a retro feed of the model and then may help to understand the flow split. The best agreement a mod el will have with the experimental data extracted from the laboratory reactor, the more useful this latter will be for the understanding of exotic fuel kinetics and pollutant formation mechanism s According to these considerations it is desired to be able to reach a loading parameter of at least 4 in a first implementation basis and to reach 10 on the long term. However, the base load or design condition is based on a


60 Loading parameter of 1. This means that the fuel flow consumption of 0.5gal/hour will be based on this regime. This gives a total air flow rate of 7g/s at this condition and for the stoichiometric mixture. Using the equation below, max /300 1.8,iT tam V PeM (0 12) t being the targeted loading parameter, and maxm the maximum flow rate available in g/s, one can evaluate the volume of a reactor in which this combustion conditions may be approached within our limitations, yielding. 60. VmL (0 13) In Table 31, the different loads and the corresponding air flow are given. One must take into account that it is necessary to pump some additional air for the cooling, which may be at least equal to the combustion air. Therefore the next step in the design would be to design the cooling system and evaluate the air flow needed. This means that the overall combustion volume targeted is an iterative process. Table 31 Combustion air flow rates Loading parameter (mol/s/liters/atm 1.8 ) 1 (design point) 4 (full load short time) 10 (full load long term) Air flow rate (%of compresso r capacity) 7g/s(5%) 28g/s(21%) 70g/s(54%)


61 Heat transfer model As mentioned before, the current reactor stands between a gas turbine combustor and a well -stirred reactor. Experimental well-stirred reactor shells are made of ceramic materials which can s ustain really high temperatures. The toroidal jet reactor developed by Neniger [11] or the spherical reactor by Longwell and Weiss [6] are some examples of this type of design. The materials used provides a good isolation and the reactor could be considere d as nearly adiabatic. Even though ceramic thermal barrier coatings have been developing a lot in the past years to protect, for example, the blades of turbines, the development of 100% ceramic combustor raises some issues. Indeed fusing and crumbling of t he ceramic shell could easily occur and give debris which would damage the turbine blades. In addition, using refractory material will limit the geometry and dimensional tolerances of the combustor which is primordial for sealing considerations. Therefore, the use of pure ceramic material for the current system is not considered and instead an actively cooled metal is used. In the absence of cooling, the pieces that would be subject to an important heat transfer from the combustion gases are the liner and t he inner tube. They will need to be cooled by a flow of air or other coolant. To design the passages and chose the material, a program modeling the heat transfer inside the combustor is developed. The program could solve either for the cooling flow required giving the input parameters given in T able 3 2 and setting the wall temperature to an appropriate value, or it could solve the wall temperatures giving the cooling flow rates. It solves a system of nonlinear equations representing the energy equation f or the metals in steady state. Assuming a one dimensional radial and steady state model, fluxes should equalize at a given wall, giving 2 equations for 2 wall temperatures per wall. The resolution is made at


62 a given location on the liner length. To obtain a temperature distribution along the liner, a step could be chosen and the cooling air temperature at a shifted position could be obtained integrating the heat flux along the length step, and using the energy equation for the gas. Table 32 Input parameters of the heat transfer model Input Parameters symbol Equation s concerned Combustion Gas flow rate gm 3.14 Fuel air ratio q 3.9 Flame temperature T g 3.11,3.14 Materials Thermal conductivity liner k l 3.15 Thermal cond uctivity tube k t 3.15 Thermal conductivity casing k c 3.15 Emissivity liner l 3.16 Emissivity tube t 3.16 Emissivity casing c 3.16 maximum temperature liner T t 3.11,3.14,3.15 m a ximum temperature tube T l 3.11,3.14,3.15 Geometry Cross section Combustion A g 3.14 Cross section Central cooling A c 3.14 Cross section Annula r cooling A a 3.14 Thickness liner t l 3.15 Thickness tube t t 3.15 Thickness casing t c 3.15 For each wall, the steady state energy equation provides the equations below: 11111212 ()()(,)0, wwwwCTRTKTT (0 14) 22222112()()(,)0,wwwwCTRTKTT (0 15)


63 where Ci and Ri are, respectively, the convective and radiation heat transfer flux received by the wall through the side i, Kij being the conductive flux directed from the i side to the j side. The three modes of heat transfer are important in gas turbine combustion. First, the gases from the combustion zone transfer heat to the wall by radiation. The rate of transfer will be highly dependent on the luminosity of the flame. A non -luminous regime will be characterized by the absence of soot particles. In that case, the gas will emit and absorb in narrow emission bands. The surface of the wall could be considered as a grey surface with a corrected emissivity compared to a blackbody. The net radiant heat flux from the gas to the wall is given by 44( ),gwgggwRTT (0 16) g the gas emissivity and g the gas absorptivity Lefebvre and Herbert investigated the emissivity and absorptivity values which are respectively dependent on gas and wall temperature, and provided the approximation [41] : 1.5.gg gwT T (0 17) In addition, v alues for the non-luminous flame absorptivity can be obtained using the formula provided by Reeves [42] 0.51.51exp290(),g bgPqlT (0 18)


64 where q is the fuel air ratio and lb the beam length. The beam length varies with the shape and the size of the gas volume. Fishenden and Saunders provided an expression for this beam length for most p ractical combu stion situations [43] : 3.4,bV l A (0 19) where V is the flame volume, and A the l imiting area between the flame and the wall. w) for the fact that the liner surface is grey, E quation 3.7 gives:,gw wgggwR TTT (0 20) w is the emissivity of the wall. In the case of luminous flames, soot particles present in the flame are very hot and emit in the visible spectrum. Particularly important in high pressure combustors, as the soot particles can get larger, this luminosity may need to be accounted in these cases using a luminosity factor as described by Lef ebvre [44] .The soot formation process in a tmospheric flames is a complex mechanism highly dependent on the atomization and vaporization processes of the fuel. Some studies have been done on the soot formation process in well-stirred reactors at high equivalence ratio by Nenn iger and Manzello among others [11,14] However, caution must be exercised in using correlations obtained from practical gas turbine combustor, which might not be pre vaporized and which geometry could be very different from our design. The mechanism of soot formation by drople t burning is very different soot regimes from that occurring in a rich premixed flame. In addition the adiabatic temperature of the combustion gases peaks at an equivalence ratio around 1.05 for the n heptane mechanism provided by Lawrence


65 Livermore National Laboratory [20, 21] Works on toroidal well stirred reactors in series with PFR showed the absence of flame generated particles in ethylene flame below an equivalence of 1.9 [14] Overall, neglecting the luminosity of the flame will not underestimate th e heat transfer rate as long as this phenomenon does not occur at the critical thermal conditions. The soot formation and oxidation mechanism will be discussed in detail in C hapter 6. The convective effects are harder to characterize in this combustor desi gn due to the complexity of the aerodynamics. The configuration of the injection is radial by opposition to the axial swirler in conventional gas turbine combustor, so that correlations used in gas turbine cooling could not really apply in this case. The m odel widely used for estimation of the Nusselt number in gas turbine combustor is based on a fully developed flow with constant surface temperature in a circular tube give the following approximation: Nu (0 21) This is valid for a range of Pr andtl number between 0.5 and 1 [45] The assumptions of constant temperature or constant heat flux may be questionable in gas turbine combustion. However the effect of peripheral heat flux variation could be accounted using heat flux functions as described in the same book. Some calculations have also been carried by Kays and Leung on fully -developed flow in c oncentric circular tube annuli [4 6 ] In the current system, a jet of fresh mixture impinges on the inner tube and mixed with the combustion gases. One can evaluate a worst case in which it is assumed that the gases are at a uniform adiab atic flame temperature in the combustor.


66 In that case, one may use the fully developed constant temperature model on a turbulent duct. The Reynolds Number will be estimated using: 22 44 Re () ()()gghg g g io gggio gioUd m m dd dd dd (0 22) and the convective heat flux from the gas to the wall could be given using Equation 3. 12: 0.8 0.50.5 ().ggg gw hskCpm Cgw TT dA (0 23) If one neglects axial heat flux, the radial conductive heat flux along the thickness of a liner or a tube is give n by: 12 12() .ww wkTT K t (0 24) On the cooling side, the radiation heat transfer can be e xpressed as the heat transfer between two gray surfaces of a given emissivity: 44 21 12 12 1 111222 1() 11 1 TT R A AFAA (0 25) where F12 is the geometric shape factor between the surface 1 and the surface 2. This one could be estimated equal to unity for a long annular space. One can extract values for the emissivity of Nickel based alloy from Greene at various tem peratures and level of oxidation [4 7 ] Machined stainless steel emissivity would be considered equal to 0.14. Finally, the convective heat flux due to the air flowing into the cooling passages could be estimated using the same constant surface temperature fully developed flow


67 described in E quation 3.15 considering the gas as cooling air and the temperature of the exterior wall of the liner or internal wall of the inner tube. The heat transfer processes neglected in this model are the radiation heat flux fr om the liner to the tube, assumed to be negligible compared to those due to the combustion gases, the convective heat flux from the ambient on the casing assuming a quiet atmosphere, and all the axial conduction heat fluxes. Figure 31 2 Representati on of the cooling passages One can notice in the equations above that the E quations 3.5 and Equation 3.6 will be non-linear. Therefore a solver which is based on a NewtonGauss method is used that is included in the function fsolve of Matlab 7.5.0. In addition, the heat transfer fluxes are integrated over the length of the combustor in order to evaluate the heat losses from the combustion volume and the temperature rise of the cooling air. Some constraints have to be considered when using this model. First if one considers the size of the cooling passages, one might be attracted to reduce them to 1/16. Indeed, if one considers only the system of E quation 3. 6 and Equation 3. 7 at a Casing Central cooling flow Combustion zone Annular cooling flow


68 single position over the length, the velocity of the cooling flow being very large, the cooling convective heat flux will be very large and the wall temperature low for a given heat flux. However using the energy equation for the gas over the length of the cooling passages, it could be noticed that the cooling air temperature will rise significantly, and that the constant temperature model will not be valid. Also, the velocity should stay under the sonic level for wall buckling limit reasons and because the correlations used at a given point for heat transfer calculations consider t he flow as incompressible. Changes of density are not neglected and have been evaluated using the gas dynamics model developed and which will be presented after. In addition, it is desired to limit the interaction between the cooling flow and the combustio n gases in the dilution zone, and so prevent the flame to attach inside the cooling passages. For this reason the momentum of the cooling flow, at the exit of the passages should not be lower than the momentum of the hot flow from the combustion zone. In addition to these aerodynamics considerations, reducing the wall thickness to the minimum could be attracted since it increases the conduction fluxes and decreases the temperature difference across the wall. The injection system designed in this project sea ls with the liner wall. The machining of the liner sealing surfaces would be very difficult on very thin walls. In addition the wall buckling limit will be lowered. Therefore, the thickness of the liner was set at 0.2 inches. The constraints are summed bel ow: 0.3 100 .a g aaggM TK VV (0 26) The choice of material is very important in heat transfer c onsiderations. Indeed, as noted in T able 32, the thermal conductivity in the conduction flux and the emissivity in


69 the radiation flux will depend on the material chosen. Other machining and cost considerations should also be taken into consideration. Ther efore, the objectives and constraints have to be described in the research of an adequate material. It is desired to maximize the maximum service temperature in air or the melting point of the metal. Increasing the service temperature will reduce the need for cooling, and so decrease the overall heat transfer. If one thinks of a global Newtons law of cooling, the heat transfer is a monotonic function of the temperature difference. The higher the metal temperature, the lower the heat losses induced by the c ooling and so the closer the system is from the adiabatic condition. The maximum service temperature does not depend only on the melting temperature, but can take into consideration other parameters as the mechanical strength at high temperature or the res istance to corrosion. The resistance to corrosion is a more complex criterion giving the different forms of corrosion mechanical failure modes. It is particularly important when injection of water is used as a NOx reducing strategy and for marine propulsio n system. The thermal conductivity may have an important role in keeping the axial component of the heat flux low, and so reducing the temperature gradient. However the direct influence on the radial heat transfer could be less than expected. The results f or a given geometry and some different heat conductivities are plotted in F igure 3 1 3 Plotted below are the wall temperature for a reference conductivity set at 10W/mK, and comparisons to the temperature obtained for two other conductivities. We can obse rve that the conductivity plays an increasing role with an increase in the cooling flow rate, and that the role declines with an increase in the heat conductivity.


70 A ) B ) Figure 313. Influence of the conductivity in the heat transfer, experimental results ( A ) and steady state wall model ( B ) The interpretation could be easily done based on a steady state wall model represented above. An electrical analogy would give: (),ga TTT q R (0 27) where RT is given by: 1 11 T ga R L hkh (0 28) The effect of the ratio of L and k is not significant unless the convective coefficient grows or the ratio of the conductivity to the length becomes really large as demonstrated in the s ensibility expression below: T1 T2 L h a, T a h g ,T g


71 21 (/) 11T agR LK L Khh (0 29) One can notice that the effect of K is the exact inverse of the material thickness when considering only the radial heat flux. Table 33 Performance functions for material choices. Functions symbol objective extreme value 1.Melting Temperature T m to maximize 1000K 1'.Maximum service temperature (air) T s to maximize 1000K 2.Thermal conductivity K to maximize 15W/mK at 298.15K 3.Resistance to Corrosion at 1000K to maximize 4.Machinability (Hardness Vickers) H v to minimize 200 at 298.15K 5.Density to minimize 15g/cc 6.Cost of material C m to minimize 15$/lb The extent to which the material is easy to machine is also very important. In the current case, the sealing surfaces being critical on the injection surface, one cannot tolerate large error s in clearance due to machining difficulties. The hardness of the material could be used to determine its machinibilit y index as described by Thiele [4 8 ] In addition, the materials involved in the heat transfer calculation do not have the same order of priority for the functions described in the T able 33 Indeed, the liner and the injector have to be machined precisely and sustain fairly high temperature. The rig casing does not see high temperature so the mechanical strength and the cost are also impor tant as it is the larger piece. The inner tube will see a very high temperature due to the jet impingement. Indeed, the velocity used to determine the convective heat transfer rate is based on the


72 total mass flow rate. However due to the turbulent jet, this velocity will be higher. Correlations for axisymmetric turbulent jets could be are given by Turns [48] and the mean velocity components respectively in the radial and axial direction are: 2 23 1, 84e xJ v x (0 30) 1/2 2 323 1 1, 16 44e r eJ v x (0 31) with Je being the initial jet momentum and 1/23 1 16e eJ r x (0 32) and the eddy viscosity is: 0.0.285, e vR (0 33) where ve is the initial jet velocity. The axial and transverse axes of the axisymmetric jet are based on the initial jet position. As described in the geometry, the rotation of the injectors between runs is possible and offers a variation into the jet direction and the impingement angle. Assuming no cross flow and no reflection, the impingement velocity could be estimated as a direct function of the distance and the direction. To do so, some coordinate systems changes can be done using a rotation matrix. For an injector angle of 10 and the direction of the jet into the tube normal plan, the velocity was estimated at 100m/s. The temperature of the jet at impingement will also be very important. However, the determination could be fairly complex due t o cross mixing with the combustion gases. It is evident that the temperature should be fairly lower than the reactor temperature of


73 2400K used in the model. In front of these unknowns, it was decided to implement first a very high temperature resistant and conductive metal. The simplicity of the inner tube does not require high precision machining and its low volume will decrease the large price of the raw material so that one can think of purer materials while controlling cost. On F igure 3 1 4 are positioned different metals with their service temperatures and their thermal conductivities. For pure material, the melting point is considered. One can observe three main groups on the graph. First, molybdenum is far apart with outstanding melting point and therm al conductivity respectively at 2617 C (2890K) and 138 W/mK (at 300K). Then one finds Platinum Zirconium and Titanium with elevated melting point but small conductivities. That justifies their use in rocket and aircraft design since they also present an excellent mechanical strength and a low density. Then, in a third group are the different alloys including Nickel based alloy Cobalt based alloy and Stainless steel. Their thermal conductivities are comparable. Inside this latter group one finds on top the nickel and cobalt -based alloys which ally high service temperature with a strong resistance to oxidation. They have been ex tensively used for decades in gas turbine combustor design. The properties of some of these alloys together with Molybdenum are presented in T able 3 -4. These data have been extracted from Matweb (www.matweb.com). It must be noticed that the thermal conduct ivity varies with the temperature, with some correlation available for Nimonic 75 and Inconel 600 [ 50 ]


74 Figure 31 4 Thermal properties of different materials Table 34 : Properties and composition of some tentative materials Properties density k ( W/mK) k (W/mK) Ts Ni Co Mo Cr (g/cc) at 20C at 800C (C) (%) (%) (%) (%) Haynes 230 8.97 8.9 28.4 1095 57 <5.0 2 22 Haynes Hastelloy X 8.22 9.7 24.8 1095 47 1.5 9 22 Haynes 188 8.98 10.4 24.8 1095 23 42 0 22 Nimonic 75 8.37 11.7 NC 1095 72 0 0 1 9.5 Molybdenum 10.22 138 112 (727) 2600 0 0 100 0 Inconel 600 8.47 14.9 1095 73 15.5 When choosing the material for this laboratory scale system, one should keep in mind that even though some properties such as mechanical strength or price may not be as important as it would be in a gas turbine combustor, ignoring them will prevent the adaptation of the system to a future industrial use. Even though thermal properties of molybdenum are excellent, it suffers from two disadvantages which are its elevated hardness (HV=230) and the stability of its oxides. Its hardness restrains the machining processes so it could not be a good candidate for the liner, but would suit the inner tube


75 requirements. Concerning the liner and the injector, copper could be a candidate if its melting temperature would be higher. It cannot t be used for temperature over 700K. Zirconium, platinum and titanium present fairly high melting temperature, platinum offering in addition a good conductivity. However the price of the pur e material is enormous. Finally, n ickel and cobalt -based alloy offers a high service temperature, a good resistance to corrosion, and a fair machinability. The choice has been made toward Hastelloy X for its excellent resistant for corrosion due to its mol ybdenum content for the liner and the injector. Concerning the tube, molybdenum is a good option in the short term. The results of the heat transfer calculations are presented in F igure 3 -1 4 First, the temperatures of the wall on the combustion sides hav e been set to a target value and the program solves for the mass flow rate at a given geometry. The central air cooling and the annular air cooling versus the flow of combustion air are plotted on F igure 314 A As expected, an increase in the combustion a ir flow will enhance the convection heat transfer so that a higher cooling flow rate is needed to keep the wall temperature constant. The different results are attached in the appendices. From a controlling point of view, the temperature of the liner could be kept nearly constant by measuring the liner wall with a thermocouple and feeding the flow controller using a (Proportional Integral Derivative) PID or another type of regulator based on both the results and a proportional integral action. In order to c haracterize a range of operation, one has to consider the heat losses in the combustion chamber. On F igure 314 B the heat losses are plotted in terms of axial bulk temperature drop along with the axial cooling air temperature rise.


76 A B Figure 3 1 5 Cooling flows required A) and axial temperature changes B )


77 These are the results of heat rate integration along the length of the liner and the tube. These results have been obtained fixing the wall temperature for different flows here expressed in terms o f loading parameters. It is seen that a bove a loading parameter of 0.5 moles. s1. liters1. atm1.8, the axial temperature changes are within a 100K range. As expected the changes are decreasing with an increase in the load. Indeed, as the convection coefficient increases the ratio of the total losses over the mass flow decreases and so does the change in bulk temperature as described in C hapte r 2 with E quation 2 -5 Gas dynamics model As defined in the overall objectives of the rig, the flexibility of the s ystem is a priority, which means that the rig should stand a large range of flow rates, gas composition and temperature. In order to achieve a sufficient mixing of the air and the fuel, and then to provide a good stirring in the reactor under the targeting range of operation, the evolution of the pressure, temperature and velocity in the supply lines and cooling passages should be estimated using a gas dynamics model. The stirring criteria in the combustion zone will be highly dependent on the injection jet velocity or pressure drop. Indeed, Kumar and co-workers investigated some scaling criteria for designing a mild burner, including keeping a pressure drop of 600mm of water gauge [5 1 ] The injectors should be designed in order to provide at least this pres sure drop over the targeted range of operation. In addition, the pressure drop along the supply lines could bring the flow to an unexpected sonic regime if it is not correctly estimated. The model developed here is a one -dimensional gas dynamics model which accounts for geometry change, singular and regular pressure drops, and heat transfer. Between each node, four consecutive objects are used to evaluate Mach number,


78 pressure, temperature and density at point n+1 giving the values of this set of variables at point n. These four objects are the isentropic flow model accounting for area change, the Fanno flow in constant area duct accounting for friction effect, the pressure drop due to the singularities of the flow, and the Rayleigh flow accounting for heat transferred to the fluid. These are shown on F igure 3-15. Figure 31 6 Gas dynamics model programming scheme The isentropic flow model assumes the gas to be perfect and the flow to be adiabatic. The Mach number of the gas fl ow will be directly related to the ratio between the section of the duct and the critical section noted A* at which the flow is exactly sonic. In this isentropic flow, the stagnation properties T0, P0 0 stays constant. Therefore one can estimate the set of properties knowing the area change and the Mach number downstream using: 111 2 21 2 21 ** 11(). AAAA Mff ffM AAAA (0 34) The pressure drop module is made of two parts. The first one representing the friction effect and the model is called Fanno flow. A one dimensional flow of constant specific heat perfect gas through a perfectly insulated constant area duct is named Isentropic flow Singularities Sudden changes in geometry Rayleigh flow Heat transfer Fanno flow Friction effects M 1 ,p 1 1,T1 M 2 ,p 2 2,T2 M i ,p i i,Ti M i ,p i i,Ti M i ,p i i,Ti


79 fanno line flow. The program here is based on the Fanno coefficient of friction. The friction makes the flow moving toward the sonic condition. If the flow is subsonic, the Mach number increa ses as one moves downstream. The stagnation and static pressure, temperature and density that the gas would reach at the sonic condition (downstream of the duct) will be noted with the superscript *. At a given position 1, one can evaluate the quantity bel ow knowing M, max 1 1(). fL fM D (0 35) in which D is the diameter and Lmax is the length downstream of this point that would be needed to raise the flow to the sonic condition. By the same way the Mach number could be evaluated at a position N+1 separated from N by the length L using the equation below, max max 21 fLfL fL DDD (0 36) The Mach number at 2 is obtained using E quation 3.27 Between these two points, there will be a loss in stagnation pressure but no loss in the stagnation temperature. The temperature increases, while the pressure and the density decrease. The friction effect could lead a deflagration travelling inside a duct toward a detonation. It is fundamental in combusti on to control this pressure drop in order to prevent damaging explosions. To use these equations above, one needs to evaluate the Fanno coefficient of friction. For laminar flow, concerning Reynolds Number less than 2000, the Fanno coefficient could be des cribed the equation: 64 .Re f (0 37)


80 For turbulent i nternal flow in a duct, with Reynolds number more than 2000, the Colebrook equation can be used: 1 2.51 2log 3.72 Rehk d ff (0 38 ) where k is the roughness of the pipe, and dh the hydraulic diameter. To solve E quation 3.29 a trial and error iterative scheme has been programmed in the Fanno module. The flow path includes some singularities; hence a n ad ditional module has been programmed to take into account their effect on the flow properties. Three types of singularities have been taken into account: Sudden widening of the duct section Sudden restriction of the duct section Sudden change of direction of the flow (elbow ) For each of these cases, a coefficient K of pressure drop is calculated based on the geometry. The equations for its calculation are given in Appendix B Knowing th ese ; one can calculate the loss of stagnation pressure through a singularity using the equation: 2 0201, 2 U PPK (0 39) where the points 1 and 2 are respec tively located upstream and downstream of the singularity and U is the mean flow velocity. Finally, the effect of heat transfer from the gases is accounted using a Rayleigh flow model. If one considers a one dimensional frictionless perfect gas in a const ant area duct, the heated flow will move the Mach number toward the sonic condition (M=1). The opposite effect occurs when one cools the flow. So for subsonic flow, the Mach


81 number increases as the flow is heated. In a Rayleigh line flow, both stagnation p ressure and stagnation temperature are affected by the heat transfer. Therefore, the superscript is used for the properties that the fluid would have at the sonic condition. The ratios between the properties at a given point and this properties are fun ctions of the Mach number. By knowing the heat flux, one can calculate the difference in stagnation temperature using the energy equation: 2 0201 1,pmCTTqDdl (0 40) where q is the heat flux per unit area from the surroundings to the fluid. Therefore, the Mach number at point 2 can be evaluated using: 111 02 0201 02 21 ** 0 010 01(). TTTT Mgg ggM TTTT (0 41) The relations between the properties of the flow for the 4 types of flows are giv en in Appendix A Finally, one can notice than in the current rig, the pressure is set by the downstream pressure as the gases exhaust to the atmosphere. It is desired to calculate the pressure downstream that would provide the desired flow rate to the com bustor. A trial and error method has been programmed for the upstream pressure that would equalize the downstream pressure to the ambient for a desired flow rate. A simulation is presented in F igure 31 7 Two peaks of Mach number can be observed correspond ing to the air nozzle for the premixing at point 2 and to the injector at point 5. The pressure drops are also significant at these two points. A small pressure recovery could be noticed as the flow is injected in the combustion zone at point 6 due to the larger section which is counterbalanced by the heat transfer from the combustion gases.


82 Figure 31 7 Pressure and Mach number evolution in the combustion air flow path Figure 31 8 Pressure drop across the injector for different injectors


83 Figur e 31 9 Jet M ach number for different injectors Using this gas dynamics model, one can design the injector that would satisfy the condition of a high pressure drop across the injection to ensure a good mixing. Simulations have been done for 4 different d iameters and are presented below in Figure 3 -1 8 and Figure 3 1 9 From the results presented, we see that with a 3/16 one may be able to cover the range 5-18g/s. In addition, it is plan to add another set of injector at 1/4 so that we can cover in the lo ng term higher range of flow rates. In this chapter, the geometry of the burner and the different models used to obtain get the design parameters were considered. The geometry investigation gave the shape of the burner and the concept of radial multiple i njections. The kinetic simulation determined the volume of the reaction zone, the heat transfer model provided with the size of the cooling passages such as the flow rate needed, and finally with the gas dynamics model the injectors were designed.


84 CHAPTER 4 FINAL DIMENSIONS OF THE RIG In the previous chapter, the geometrical concept of the burner was provided and the different geometrical parameters were determined using different simple simulation approaches. From the results of the simulations and based on some criteria established, appropriate dimensions were selected. The final values of these dimensions were also determined from an economic viewpoint, which means that standard tubing and standard drill sizes were used as much as possible. The final des ign is presented on the next pages on F igure 4 1 and Figure 4 2. The rig is made of 7 main pieces whose quantities and material is presented in the T able 41. The liner is actually originally made of two pieces, which is the cylindrical body and the swirling vane, which is an annular piece, tapped and then pressed into the main body. Table 41 List of the main parts of the combustion rig Number Piece Material Quantity 1 Liner Haynes Hastelloy X 1 2 Injector Haynes Hastelloy X 4 3 Casing Stainless steel 304 1 4 Inner liner Molybdenum 1 5 Tube Stainless steel 304 1 6 Cooling air liner Stainless steel 304 1 7 Distribution plate Stainless steel 304 1


85 Figure 41 Axial view of the rig (Scale1/1) Combustible mixture Annular cooling air Annular cooling air Central cooling air Exhaust


86 Figure 42. Normal vi ew of the rig (resized)


87 The figures above being at scale 1/1, the dimensions could be read directly from the figure. Concerning the volume of the combustor, this is an annular shape of dimensions: 1.08 (27.4mm) of internal diameter, 2.076 (52.7mm) of exte rnal diameter and 1.8 (45.7mm) of length. Therefore, the total volume of the primary and secondary reaction zone (see F igure 3-5) is about 4.5 cubic inches or 75mL. One can notice that a rod has been mounted inside the molybdenum tube to increase the cooling flow velocity. The cooling flow path enters the rig trough a central hole in the distribution plate shown in F igure 4 4, and flows to a prechamber in which the liner is tapped 20 times. The function of this large volume chamber is to equalize the flo w and therefore provide a good distribution around the circumference. The air flows radially and then cools the rear part of the liner before entering the molybdenum tube at the end mixing with the combustion gases. Concerning the annular flow, the air is provided through 4 holes in the distribution plate ( F igure 4 4). It enters an annular volume of large section before flowing through 50 holes of 1/16 diameter to enter through cooling section of the liner presenting a 0.08 clearance. The 50 holes are set at 20 degree angles with the axis of the liner and in the tangential plan. The function of these holes is to provide an important pressure drop between the annular pre -chamber and cooling passages so that a good distribution of the flow is obtained. In addition the angle provides a swirling effect, with the effect to increase the absolute velocity in the cooling passages.


88 A B Figure 43. Rig assembled, outside A ), and combustion zone B) A B Figure 44. Back view of the rig with distribution plate A) and without B ) The critical parts in the current design are the injectors. The radial injection includes a slight 10 angle, hence designing a shape of an injector which would not signific antly Injector Optical port Swirling Central air Central liner


89 perturb the flow was a challenge. A previous tentative design with no injection angle and intrusion into the injection chamber is compared to the current design on F igure 4-3. In the final design the injector is a cylindrical part with a neck for coo ling reasons. The sealing is made with a plane to plane contact and an axial constraining stress was used with a flange lock nut pressing a counter sink on the casing. A B C Figure 45 Injector design: previous invasive design A) and final design with normal B) and C) isometric views If one considers the small size of the liner diameter, one can figure out how a high precision machining will be critical in the performance of the combustor. The clearance between the liner and the injector was specified at 0.005 for the smaller diameter. Figure 4 3 shows that in addition there are injector ports on both the liner and the casing so that geometric tolerances should be accurately addressed. Indeed, the cylindricity will provide a reduced clearance for additional sealing and a possible rotation for flexibility. The coaxiality of holes and their perpendicularity with the countersink on the liner and the rig are essential to the main plane on-plane sealing. To get good sealing su rfaces, the casing and the liner have been assembled during the final machining and corrections have been made to match the clearance requirements. Considering the Sealing contact Fl ange locknut


90 geometrical constraints and the discussion in C hapter 3 concerning the pressure drop across an injector, it was decided to machine a set of blank injectors. This set of injectors has the same functional surfaces with the liner and the casing than the tapped one, but no hole has been drilled. The latter has been based on a 1/4 outside diameter bu t the exact diameter can be decided based on the preliminary experimental observations. In addition, given the small size of the assembly, the threads are mainly British fine thread. The complete machine drawings with the specifications are provided in A ppendix B. A B Figure 46 Injector ports on rig casing and liner A) and injector B ) The machining operations have been performed by T.M.R. Given the complexity and the precision required for this combustion system, a continuous contact has been maintained with the machinist supervisor. All the operations have been done manually. The tool and the rotation speed are given in T able 42. No real difficulties have been encountered on the different parts. Molydbdenum needs a soft tungsten carbide due to its ten dency to crack and chip. For Haynes Hastelloy X, no more than 0.05 could be cut in one pass. Very good sealing have been obtained on the injector ports which was


91 critical in this design. In addition, a very good coaxiality could also be observed for the molybdenum tube, the rod and the liner which was challenging given their mounting scheme. On an economic point of view, computer manipulated control of the drilling and turning operation would be viable starting at a series of 4 rig assemblies. Table 42 Machining parameters used: Material Tool Rotation speed (rough Stainless steel 304 C5 Carbide 500rpm Haynes Hastelloy X C2 Carbide 220rpm Molybdenum C1&C2 Carbide 220rpm


92 CHAPTER 5 DESIGN OF THE DELIVE RY SYSTEM Air supply The air supply should be designed in order to provide a combustible mixture whose composition, temperature and pressure is accurately controlle d In addition for premixed combustion, the fuel and air should be mixed before entering the combustion zone As mentioned before, premixed combustion raised the problem of the presence of a flammable mixture downstream of the injector. Auto ignition and f lashback in unexpected sections of the system could be very damag ing This is particularly true in flow controllers which are expensive material and cannot stand temperatures above 70C As summary of the conditions that the delivery system has to accomplish the list below has been established: Delivery of the combustion air flow to the four injectors (2 20g/s) Delivery of the cooling air to the cooling passages Heating of the injection air to the desired temperature (400 -700K) Atomization, Vaporization and mixing of the fuel in the air mixture Precise control over the four injector mixture composition. Compressor, dryer, filters and receiving tank A compressor pumps the air from outside to pressurize a tank. The maximum pressure the tank could support is 175psi. The compressor has been chosen according to the maximum power available on the laboratory site corresponding to 7.5hp. The pressurized air could serve for other activities in the laboratory such as air powered machine tools. The idea here is to work on a duty cycle. Indeed, the compressor will be turned on as soon as the pressure in the tank is lower than a specified value pressure. When the pressure is reached the compressor will be turned off. During operation of the rig, air will be drawn from the tank on a continuous basis. Therefore, the pressure will


93 keep decreasing in the tank until a specified minimum pressure is reached and then the compressor will be turned on. When selecting the compressor technology, consideration was given to the cost but also the fact that the compressor will have a discontinuous duty. Large current are drawn from the network when starting a compressor and so electrical starters are employed. For these reasons, a splash lubricating reciprocating engine was selected A more silent apparatus such as rotary screw compressor are more expensive and not designed to be start ed many times per day. For a reciprocating engine, t he compressor and tank were investigated under various loads from the combustion system in order to m inimize the starts and to calculate the switch pressures. The compressor could be model ed as a volumetric pump which means that the volumetric flow rate is constant and the pressure determined by the receiving tank pressure. The tank could be represented as a constant volume system undergoing change in pressure. An explicit Euler scheme could be used to represent the pressure evolution in a tank described below () () ,inout tt TmmRT PtPt V (0 42) () (),inPt mVt RT (0 43) w ith VT being the total volume of the injector and V (t) being equal to the volumetric flow rate of the pump when turned on or zero if turned off. A simulation of the pressure evolution in the tank i s show n on F igure 5 1 which show s that the compressor could be turned on only 8 times when 35SCFM are drawn to the combustor assuming a 200 gallons tank. The manufacturer advises to keep an average amount of start s below 6


94 starts per hours on a continuous operation. The maximum pressure in the tank is set to 180 psi which is the tank constructor rating and the minimal pressure at 80psi so that the equipment downstream does not see too large of pressure variations. A dryer is used in line with the compres sor to remove moisture and cool down to 50F along with two high efficiency filters to remove small particles. Figure 51 Pressure evolution in the tank at 35 SCFM Table 51. List of products for the air supply Product Supplier Description QT 7.5 7.5.80 Quincy Simplex air reciprocating compressor 24.2 CFM at 175psi, mounted on an 80 gal. vertical tank QPHT 50 Quincy High inlet temperature refrigerated air dryer 40F 40SCFM with aftercooler and coalescing air filter. CPNT 00030 Quincy Standard coalescing filters at 30SCFM and 100psig, 99% at 0.02 micron and with pressure difference gauge. 5018260 200 Quincy 120 gallons receiver tank


95 Air Heaters Design specifications for the combustion system include preheating the combustion air to a temper ature up to 700K for partial load (7g/s) and 400K for all loads as mentioned in the C hapter 2. With t he four injectors being separately controlled, and the flow controllers not being able to with stand these temperatures, four independent heaters need to be used. The energy equation could be used to estimate the power required: pinoutmCTTQ (0 44) Based on a 283K inlet temperature (from the dryer), the required load at 700K requires 732W and the full load at 400K, 587W. The heaters are controlled on an ON/OFF basis using an auto tunable PID CN132 from Omega. This latter items considers the difference between the set point and the signal of a type K thermocouple at the exit of the heaters. This controller being autotunable, it sets automatically the 3 constants of the proportional integral derivation action Fuel and air mixing system The reference fuel, C7H16 is mixed with the air upstream of the four injectors. The objective in this part is to provide a good control and mixing of the fuel and air mixture. That means the goal is to control the fuel -to air ratio independently for the four i njectors and to minimize the radial distribution of the mixture composition at the injection. Recall that, p remixed combustion offer s large advantages in NOx reduction. However, this regime is difficult to achieve with liquid fuel. In gas turbine combustion, auto ignition of the fuel limits the residence time in the premixing section so that the processes of


96 atomization, vaporization and mixing should be quickly done. The mixture should be as homogeneous as possible to limit the fuel rich zone, with higher flame temperature and therefore higher NOx formation. As this rig aims at being very flexible, runs should be possible over a wide range of flow and inlet conditions. Therefore, the range of fuel flow of interest has to be defined As considered before, the minimal air flow rate that would lead to an adequate mixing could be estimated at 3.5g/s. The lowest fuel rate would so be obtained at the leanest flammable mixture at this flow rate. Weiss and co workers investigated the limits of heptane in a well st irred reactor and provided with a lean blowout equivalence ratio of 0.5 for a loading parameter equivalent to the current 3.5g/s and 0.6 at 20g/s (Weiss et al 1958). As noted in the C hapter 3, given the geometry of the set of injectors, the system is limited concerning the maximum flow rate of air. Knowing the air flow rate, the mass flow rate could be calculated directly on the equivalence ratio basis: fa stochmmF/A (0 45) From E quation 1, the lowest flow of fuel can be estimated at 0.11g/s .Concerning the highest limit, the mass flow of air is limited to 20g/s. This is the limit the firs t set of injector s but a second set gives the opportunity to go up to 30 g/s. The equivalence ratio is limited for the highest flow to the stoichiometry. Table 5 2 Range of total air and fuel flow rates Air flow rate Lower Fuel flow Upper fuel flow 3.5g/s (LP=0.5) 0.11g/s 0.40g/s 20g/s (LP=3) 0.76g/s 1.3g/s


97 It was previously discussed that the air temperature would be set around 400K and up to 700K in the diluted regime. The objective is to get complete vaporization of the fuel in the prem ixing section so that the mixture could be considered as homogeneous at the injection. The boiling point of the reference fuel n-heptane is 371.58K. One could achieve complete vaporization of the fuel in a heat exchanger separately but then the power to be transferred to the fuel by the heater would be very large which may require a very large heat exchange area. In order to still favor vaporization, the fuel could be heated upstream of the premixing point so that the total heat that has to be transferred from the air to the fuel to reach the evaporation curve is lowered and the mixing length could be diminished. In th e current design, a liquid fuel heater is designed to raise the fuel temperature just below the boiling point. The range of fuel flow rate being wide, a controller has to set the temperature of the fuel to the desired value. As the four injectors are independently controlled and the flow metered, 4 heaters have to be designed. Indeed the peristaltic pump could not stand 370K due to their pol ypropylene fittings so the pumping and metering of the flow has to be done upstream of the heating system. The heaters are made of a tube coated with a flexible heater. Two parameters have to be chosen, the diameter and the length of the tube. In order t o design the heat exchanger, we use a constant surface temperature model for internal flow is used In fact a flexible heater in a duty cycle controlled by a P.I.D action is used to keep the liquid exit temperature at the specified value. The heat transf er at a given section of the heater is given by: ()()().conv slqxhTxTx (0 46)


98 Integrating the heat transfer rate over th e entire length of the heater one get s : 2 1() exp, ()Sl Sl pTT PLh TTmC (0 47) where 1 and 2 are respectively the inlet and outlet liquid temperature. Inver t ing this relation to get Ts as a function of L gives: 21. 1p pPLh mC ll s PLh mCTTe T e (0 48) Using the energy equation and assuming steady state for the metal tube, one get s : convplQmCT (0 49) To be able to work in duty cycle and also fix the surface temperature, one want s that the power transferred to the fluid be less than the maximum power of the heater .convheaterQP (0 50) In our case, the maximum Qconv needed to raise the maxim um flow r ate to the boiling point is 150W per heater. The area covered by the heater tape for is 149 cm2. The above equation is then used to calculate the surface temperature that this system would need. To get an estimat e of the heat transfer coefficient, the correlation of the Nusselt number defined for an internal flow is used as: hk Nu D (0 51)


99 Considering the low flows under investigation (down to 1mL/min), it is crucial to perform a proper dimensional analysis and so be able to use the relevant correlation. Some non-dimensional numbers are of importance when qualifying a flow of fluid. The Reynolds number is the ratio of the inert ia forces to the viscous forces, .eUD R (0 52) The Grashof number is the ratio of the buoyancy forces to the viscous forces 2 () ,s gTT Gr (0 53) The Prandtl number measures the ratio of the viscous diffusion to the thermal diffusion Pr .pC k (0 54) At very low Reynolds number the regime is referred as creeping flow, and the inertia term becomes negligible compared to the viscous effects. In the case of a heating duct, free convection due to buo yancy forces will appear as a consequence of the temperature gradient between the hot duct and the cold fluid. Mori and Futagami examined heat transfer on horizontal heated tube s with air [5 2 ] They found that the Nusselt number started to increase compared to forced convection when the product of the three numbers above exceeds 103. Then it increases by 50% with a 105 increase of the latter product. On T able 5 -3 are presented the values of the different non dimensional numbers. In the current work, the Gr .Re.Pr product is well above this limit so one can consider that the convection coefficient will be define d in a laminar regime.

PAGE 100

100 Indeed for a fully developed flow with constant temperature, Nu=3.56. For a 3/8 diameter, the length of heated tube will be 50cm and for a 3/16 it will be 100 cm. Using E quation 5 8 one can calculate the needed tube temperature for a heating of 150W to be transferred to the liquid. For an inlet temperature of 293K, a surface temperature of 394K for the 3/8 tube and 376K for a 3/16 tube, are obtained As 3/8 needs less length, provides an enhanced heat transfer coefficient (higher Gr Pr Re product) and a reasonable surface temperature, it was adopted. Table 5 3 Non -dimensional numbers value for the creeping flow discuss ion Tube Re Gr Gr Pr Re (10 7 ) GrRe 2 3/16 8.65 346 1.49 10 6 3.5 142 12.5 20000 3/8 4.32 173 1.19 10 7 14.2 568 400 640000 In prevaporized, premixed system, the objective is to get a homogeneous gaseous mixture of fuel and air. As one will not reach vaporization of the fuel before the premixer, it is desired to spray the liquid fuel jet into the air stream so that atomization takes place. The atomization process in fuel injection is the one consisting of disrupting the consolidating surface tension by the action of a n aerodynamic force. Indeed the surface tension is an effect of the intermolecular forces on the surface of a liquid and it tends to reduce the surface area of a liquid sheet or droplet. The minimum surface area/volume ratio being a sphere, this explains why the droplet from a faucet tends to be spherical. Using a model of a drop let breakup, the two forces should be equal so 221 (/4), 2 ARDUCDD (0 55)

PAGE 101

101 where CD is the drag coefficient of the drop let, the surface tension, D the droplet diameter, and UR the relative velocity between the droplet and the air flow Atomization is generally seen as a sequence of two processes, the primary atomization which breaks the fuel sheet or jet into shred s and liga ments and the secondary atomization which leads to smaller droplets. Two mechanisms have been distinguished to describe the fuel jet breakup process depending on the intensity of the disruptive force, which is a direct function of the intensity of the relative velocity between the air and the fuel bulk velocity. In the classical mechanism investigated by Rayleigh, small disturbances creates waves which form a neck that which further lead to breakup due to surface tension effects. In this mechanism, the dr oplet size is of the order of the jet diameter and time scale of atomization remains large. When the relative velocity is further increased, the prompt atomization regime is reached in which breakup takes place very rapidly and the final droplet size has no dependence with the jet initial size. This latter regime will be of primary interest for this work. This regime will be achieved if the jet of air is of sufficiently high velocity to promptly atomize the fuel jet and this parameter will be critical in the design. Figure 52. Mixing system

PAGE 102

102 The system has to be able to guarantee the homogeneity of the air/fuel mixture over the entire range of flow rate. Using the one-dimensional gas dynamics model presented in C hapter 5, the diameter s of the differe nt passages were chosen that would accomplish a strong atomization, but also provide the necessary mixing and vaporization. Below are listed the important parameter s and their function: 1. Air nozzle diameter: determines air jet velocity and so the atomiz ation process (E quation 5.14). 2. Fuel nozzle diameter: determines the fuel jet penetration and the rate of atomization 3. Premixing section diameter: determines the rate of mixing by the turbulence intensity The air nozzle diameter will be designed to provide the largest velocity without reaching a sonic flow anywhere in the air supply. In fact, the viscous effect on the air due to section narrowing will push the Mach number toward 1 since it could be consider ed as a Fanno flow. If the pressure drop is t oo great across an element this may cause a sonic condition downstream at this injection. Given the range of mass flow of air targeted, one must yield: maxmax((,))1,iMmi (0 56) where i designates any cross section of the flow. This shows the interest in programming a one dimensional g as d ynamics model described in C hapter 3. The fuel nozzle is limited to the smallest tube possible. Indeed, the low flow of fuel involved in a continu ous delivery by opposition to IC E (internal Combustion Engine) leads to very small velocity of the fuel. The maximum velocity will so be obtained with the smallest tube adaptable to the assembly which is a 1/16

PAGE 103

103 stainless tube with a thick wall. Heating of the fuel upstream will lead to fairly high temperature, preventing the use of Polyethylene tube. The premixing section diameter consisting of a flexible hose is also of critical importance. In fact the amount of mixing and the residence time in the premixing section will guarantee the homogeneity of the injected mixture even though there is a poor fuel jet penetration. It is desirable to satisfy the following cond ition: .mixingres (0 57) The mixing could be estimated trough the turbulence intensity. A rule of thumb for turbulent developed flow (ReD greater than 4000) in a duct consists of: 1/8 () Re,rms D u U (0 58) where U and urms are respectively the root mean square of the fluctuating component and U the time average of the velocity Then, one can very roughly approximately use a mixing time scale as the ratio between the diameter of the tube and the fluctuating component of the velocity. Indeed, the turbulent mixing wanted would be the radial one. The residence time is the length divided by the bulk velocity so that the following is maximize d : 1/8Remixing D resu L D U DL (0 59) The maxim um turbulence intensity would be achieved, in a developed flow for the smallest Reynolds Number in the turbulent region. If Re=4000, the turbulence intensity reaches 35.4%. For a first implementation a 6 length and 3/8 tube is used between

PAGE 104

104 the mixing point and the injection. This tube could be changed afterwards, in order to study some p artially or very well mixed premixing regime. The ranges of the different parameters discussed above are presented in T able 5-4 In T able 55, one can find the list of devices used in the fuel preparation/mixing system. 6 auto -tunable PID controllers are u sed for temperature control, four for the lines of air, one controlling simultaneously the four liquid fuel heaters, and 1 controlling the water heater for water steam injection fairly close to the liquid fuel heater. Table 54 Parameters for the mixing system: Condition Dimension Value (d in ) Lower flow Upper flow Air velocity Air tube diameter 0.20 49.6m/s 261.1m/s Fuel velocity Fuel tube diameter 0.0035 0.02m/s 0.25m/s Turbulence intensity in mixing tube Premixing tube diameter 0.25 32.6% 26 .3% Table 55 List of material used in the air/fuel mixing system: Device Supplier Model Description Air Heater Omega AHP 7561 4 T Type heaters (750W) Thermocouple air Omega KQSS 116G 6 4 Type K thermocouples Peristaltic pump Fisher Scientific 13 876 2 5 Medium flow peristaltic pumps(0.885mL/min) Flexible Heater Liquid fuel Omega STH051 020 5 Heavy insulated heater tapes (156W) Thermocouple fuel Omega TMQSS 040G 6 Type T thermocouple Heater controller Omega CN132 6 PID Controllers Controlle r relay Omega SSRL240DC10 6 Solid state relays

PAGE 105

105 Control over the gas lines In C hapter 2, the objectives and need for an accurate control over the injected mixture is discussed as well as a flexibility allowing staging and study of dissymmetry influences on stability. Indeed, it was planned to separately control the composition of the four injectors. The first reason is the ability to run on a staged combustion regime on the long term where two injectors run rich and two injectors run lean. In that case t he two rich injectors can be oriented toward the back of the combustion zone and the two other toward the front of the combustor. In addition, one can study the effect of a perturbation of one injector on the global reaction regime especially at low residence times. The last reason is the flow of air and fuel being fairly small; it would be difficult to have the composition fairly equal between the four injectors using some manual vanes. This is particularly true for the low rate of fuel flow so that usin g a manifold will not provide a fair accuracy on both the global and local equivalence ratio. I mplementing a certain number of different gaseous fuels is considered in this study, which are methane, hydrogen and carbon monoxide such as running in a dilut ed regime with mixtures of carbon dioxide, nitrogen and water vapor. The controllers will have to be chosen so that one can minimize the total number of controller s, and give at the same time the maximum flexibility over the composition of the four injecto rs. The gas mass flow controllers chosen for this project are calibrated for all the different pure gases presented above. A manual switch could move from one calibration to the other. For this reason, one could have a very large number of control strategies with a few flow controllers in order to optimize the flow range and also the accuracy. On T able 56 some fuel implementations of interest are shown along with the different control

PAGE 106

106 strategies depending on the operation range. The diagrams of the instrumentation are presented on F igure 5 4 Table 56 Range of gases and control configurations, 4 injectors independently controlled (4), 2 injectors independently controlled (2), global mixture controlled (1). The combustion air is assumed to be always in l evel 4. Fluid Function Control Flows Flow rates Controllers Air Combustion 4 2 30g/s 1500SLPM 4 MC500SLPM Cooling central N/A 0 10g/s 500SLPM 1 MC500SLPM Cooling annular N/A 0 20g/s 1000SLPM 1 MC1000SLPM Methane Fuel 2 0 1.5g/s 150SLPM 2 MC50SLPM 2 MC250SLPM Fuel 4 0 0.5g/s 50SLPM 2 MC50SLPM 2 MC250SLPM Carbon monoxide Dilution 1 0 3g/s 50SLPM 1 MC50SLPM Fuel (H2 0%) 4 0 12g/s 200SLPM 2 MC250SLPM 2 MC50SLPM 2 0 18g/s 600SLPM 2MC250SLPM 2 MC50SLPM Fuel (H2 50%) 2 0 6g/s 100SLPM 2 MC50SLP M Fuel (H2 5%) 2 0 15g/s 500SLPM 2 MC250SLPM M ixed (H2: 5 50%) 1 0 18g/s 300SLPM 1 MC50SLPM 1 MC250SLPM Hydrogen Fuel (H2 100%) 2 0 0.85g/s 600SLPM 2 MC250SLPM 2 MC50SLPM 4 0 0.28g/s 200SLPM 2 MC250SLPM 2 MC50SLPM Fuel (H2 50%) 2 0 0.7g/s 500SLPM 2 MC250SLPM Fuel (H2 5%) 2 0 0.14g/s 100SLPM 2 MC50SLPM Staged or mixed 2 0 0.42g/s 300SLPM 1 MC250SLPM 1 MC50SLPM Carbon dioxide Dilution 1 0 0.8g/s 50SLPM 1 MC50SLPM 1 0 4g/s 250SLPM 1 MC250SLPM Nitrogen Dilution 1 0 10g/s 500SLPM 1 MC500SLPM W ater vapor Dilution 1 1 1.5 0 22sccpm 1 Peristaltic pump N heptane Fuel 1 0 1.5 0 175sccpm 4 Peristaltic pumps

PAGE 107

107 Figure 53 Air supply Process diagram Figure 54 Process diagram of liquid fuel

PAGE 108

108 A B C Figure 55 Process diagrams of gas in le vel 4 A ), 2 B ) and 1 C ) All these regime s are possible with the 12 controllers present in T able 56. Three degrees of control are presented above. The lowest level is 1 s in which the control is only made over the total flow injected and so the four injectors present the same composition. For the level 2, the injectors are controlled two by two, this means that two compositions of the opposed injectors will have to be identical, but will be independent of the composition of the two others. The highest level is noted 4, since the four

PAGE 109

109 injectors are independently controlled and so offer the highest accuracy over the combustor composition. Concerning the study of vitiated combustion, there are different plans. First, one can dilute with pure nitrogen to obtain an oxygen rarefied combustible mixture. Indeed, with 500 SLPM of nitrogen, one can get up to 91% of dilution at LP=1 and 85% at LP=0.5. There is also particularly an interest in characterizing the effect of CO2 in the flameless combustion regime. Indeed, it is desired to investigate both the thermal effect s and the chemical effect s of dilution. Here one can achieve from 0 to 250 SLPM with two ranges of accuracy for CO2. According to equilibrium calculation s using Cantera and the n heptane mechanism describ ed in C hapter 3, stoichiometric air -n heptane mixture for 0.07g/s of total combustion air flow would create 20 SLPM of CO2 in the exhaust for an open cycle. One can see here the large range of CO2 dilution available here. Water injection is also of interes t as it might play a primary role in flameless combustion chemistry by increasing the amount of highly reactive OH radical s. Table 57 List of flow controllers Item from Alicat Scientific Quantity. Accuracy (reading/full scale) MCR 1500SLPM 1 0.8%/0.4% MCR 500SLPM 1 0.8%/0.4% MCR 500SLPM 5 0.4%/0.2% MCR 50SLPM 3 0.4%/0.2% MCR 250SLPM, 2 0.4%/0.2% Total 12

PAGE 110

110 As mentioned in the statement of scope, we are interested in running some sort of synthesis gases as well as hydrogen. In T able 5-7, are prese nted the flow needed to run different synthesis gas at stoichiometry for several flow rates of air. In shaded are the mixtures that could be implemented within at least one of the control configurations of T able 58 Zhang et al investigated laminar flame speed s for CO/H2/air mixtures, presenting different CO -to -H2 ratios, but given a constant adiabatic temperature of 1500 K [5]. Flame speeds were measured at 1.6 atmospheres and 300K of reactant temperature. It appears than from a pure H2 to a low H2/CO (5% H2 in mass), the chemical time, defined as the ratio of the flame thickness to the flame speed go es from 0.5ms to 5ms. Running this low mixture will be a challenge for the stability of the combustor hence interest increases in investigating them and comp aring the residence time at blowout for different hydrogen content s In addition running pure hydrogen is also considered Indeed, hydrogen present s a fairly simple mechanism (12reactions) and it could be used for the combustor calibration in parallel wit h a CFD simulation. Indeed, Conaire et al made a recent comprehensive study of hydrogen oxidation for conditions ranging from 0.05 to 87 bars and from 298 to 2700K and provided with a 19 reactions -mechanism [53] A reactor network model could be developed with a higher resolutio n due to the larger number of CST R reaction zone that could be modeled within a reasonable computation time in comparison to the GRI 3.0 mechanism.

PAGE 111

111 Table 58 Flow rates of H2, CO a nd global combustible mixture (air and fuel) for stoichiometric mixtures of CO/H2/Air containing from 5-50% of H2 in mass of fuel. The green shad ed area corresponds to the conditions that c an be achieved with a level 2 of control, and in blue, those achieve d with a level 1. The areas not shaded correspond to conditions that cannot be implemented using the controllers presented above. Gases in SLPM ( H 2 ( in mass of fuel) 5% 20% 50% Gases H 2 CO A+F H 2 CO A+F H 2 CO A+F f a =3.5 g/s 95 130 401 210 60 445 276 20 471 f a =7.0g/s 191 261 801 420 120 890 552 40 946 f a =20.0g/s 545 744 2289 1200 343 2543 1577 114 2696

PAGE 112

112 CHAPTER 6 DI AGNOSTICS AND FUTURE IMPLEMENTATIONS The combustion system designed in this thesis is a flexible research tool for different types of experimental studies. The implementation of diagnostics to investigate the combustion alternative fuel as well as vitiate d combustion is a long term process and many different steps may be considered In a first iteration the techniques that would help us to understand the combustion behavior of the burner are implemented to help characterize this novel design The combust ion rig, presented in C hapter 4, possesses different in situ measurements ports. An optical line of sight is created by two opposing windows in the vicinity of the combustion zone exit. These ports can also be used for any kind of cooled extractive probe, or direct flame measurement by a Platinum/Rhodium thermocouple. In addition, five ports have been incorporated to probe the liner temperature at different axial location s, and one to probe the air temperature at the exit of the cooling passages. A B Figure 61. Picture of the thermocouple ports A ) and an optical access port B) Optical port T.C Ports

PAGE 113

113 Combustor characterization The first experimental investigation planned concerns the stability loop of this system. This parameter influences directly the fuel flexibility performance of this burner. Different mixtures of air and n heptane will be tested in order to characterize the stability limits for a range of flow rates going from 3.5 to 20g/s. The mixture will be ignited with a spark or glow plug implemented through a port in the exhaust tube. A sudden drop of the liner temperature occurs when the flame blows out. Therefore, the outer surface of the liner will be probed by 5 type -T thermocouples. In addition, one can observe the flame right at the exit of the primary zo ne through the UV windows. This will tell if a luminous reaction occurs outside of the primary zone. To get information on the composition the combustion intermediates and products, a commercial portable gas analyzer probe Lancom III currently used in a research micro -gas turbine system can also probe the exhaust gases in this system. This gas analyzer can probe CO, NO, NO2, SO2 and H2S using electrochemical cells, CO2 by infrared spectroscopy and hydrocarbons using a pellistor catalyser The electrochemical cells offer a resolution of 1 ppm for the corresponding species, and the volume fraction of CO2 is assumed to be predicted with an accuracy of 0.1%. The flue gas temperature can also be probed from 0 to 1000C with an error of 5C. An additional port in the rig has been implemented so that a cooling thermocouple could probe the cooling air leaving the annular space between the liner and the casing. The temperature rise of the cooling air can provide an estimat e of the heat transfer from the liner to t he cooling air and so estimate part of the heat losses from the combustion zone. An investigation of the mixing in this design can also be performed under a cold flow regime. Nenniger used a Plexiglas version of the toroidal jet stirred reactor fed with

PAGE 114

114 water to characterize the cold flow aerodynamics [11] Dye was seeded on the water so that the flow patterns could be visualized. Cold flow techniques on air can also be used to characterize a combustor and has been implemented on actual combustion systems Madhiyanon measured mean and fluctuating absolute velocities in a vortex combustor firing rice husks, using a hot wire anemometer under cold flow conditions [ 54 ] In the current case, since this system is pre vaporized, a cold flow regime can be reached with air only and a hot wire anemometer can provide estimation of the turbulence intensity in the combustor and give an idea of the amount of mixing present. Species and Temperature characterization in the primary zone In C hapter 3, a first model of the primary zone of this reactor has been estab lished using a sequence of well-stirred reactors. Based on a detailed mechanism for n heptane, and assuming a given flow split, the product composition and temperature were simulated. The analysis of the products formed in the primary zone, using gas sampling coupled with a commercial gas analyzer, or, if possible, with a gas chromatograph-mass spectrometer will provide experimental results to be compared with this reactor model. In addition, experimental stability limits can also be compared to those of a reactor network model. A comparison will not lead to the absolute flow split and reac tor zones, since an infinite number of reaction zones would be needed for that, but is essential for kinetic investigation of different fuels and levels of dilution. The technical challenge is this part is to be able to characterize the temperature and com position of the combustion gases at the primary zone exit temperature. Indeed, temperatures above 1500K are expected at this point, as predicted by the reactor network model developed in C hapter 3. At this level of temperature, direct probing could be done using a thin Pl/Pl/10%Rh (type S ) thermocouple probe with a non -catalytic

PAGE 115

115 coating made of a mixture of YCl3 and Be0 [55] Radiation effects have also to be accounted and corrected as described by Kaskan [56] Different highly resolved laser -based techniqu es have been developed for flame temperature measurements Most of them are based on non-linear wave mixing. Coherent anti -Raman stokes s pectroscopy is based on the complex interaction between two incident waves and a free vibrational mode of the molecule probed, so that a strong resonant signal is generated at the anti S tokes frequency. In this technique, two lasers are used: a fixed frequency w1 laser such as a Nd YAG, and another frequency tunable dye laser at w2, such that resonance could be obtained wh en 2-1= 21, where 21 is the frequency of the vibrational mode. This technique is complex and requir es a strong spatial and temporal cohe rence. However, it can provide very accurate information on both major species concentrations and temperature. In additi on, as the wave generated in strongly monochromatic, line of sight implementation is commonly employed. However, it is also limited to concentr ations of the order of 1000ppm [57] Another wave interaction technique has been developed: the degenerate four -w ave mixing (DFWM). This technique can provide minor species concentration s and requires fairly simple optical implementation. Finally, spontaneous emission phenomena are also widely used in combustion diagnostics. It involves the emission of a photon from an upper energy level. Different sources could raise the molecule to an upper electronic energy level such as collisions with photon, electron, or another molecule in the case chemical reaction. In chemiluminescence, a chemical reaction results in additio n to the products formation in the emission of a photon. Indeed, a molecule is formed in an excited electronic state,

PAGE 116

116 which then decays, and emits a photon. Chemiluminescence signals of OH*, C2 *, or CH*are widely used for i n situ diagnostics to detect ign ition delay s for example. The reaction primarily responsible of OH* formation in hydrocarbon oxidation is: 2. CHOOHCO (0 60) The light could be collected and filtered around the specific fluorescence wavelength and so its time-resolved intensity could provide information on the chemical rates and the heat release. Even though the spontan eous fluorescence signal is non directive so that it emits into optical port could tell about the eventual reactions taking place at the exit of the primary zone. When the excitation is produce d b y a laser, the term of laser -induced fluorescence (LI F) is used. In this case, the energy level reached will depend on the wavelength of the incident light. Excitation in the visible or UV range, which requires the simplest optical devices compared with the Infrared region, could provide one with major and minor diatomic species in combustion, such as CH, OH, C2, CN, NH, and NO. Spectroscopic data of these mole cules are provided by Eckbreth [57 ] These advanced optical techniqu es have to be studied in detai l in terms of limits of detection, sensitivity, and sources of noise. Some more straightforward incoherent techniques could be quickly implemented in the system. In fact, the line of sight optical access enables extinction measurements, spontaneous Raman and elastic light backscattering. In Raman scattering, the incident light interact s with the vibrational mode of the particle, and a photon is emitted as a shifted frequency known as the Raman shift. Different hydrocarbons present resonance peaks in the R aman spectr um as described

PAGE 117

117 by Zhang and co-workers [ 58] However the weakness of the Raman signal requires a high fluence laser as excitation source. Some simple incoherent optical tec hnique such as extinction, back scattering, and 90 scattering techniques are based on the elastic light scattering theory developed first by Rayleigh for small particles and then generalized by Mie. When an electromagnetic 0 encounters an obstacle such as a particle, it interacts with the particle electric charges creating a dipole momentum that oscillates so electromagnetic radiation is generated. The interaction could be elastic, where the emitted wavelength is also 0 (Rayleigh scattering), or be inelastic, where the incident EM wave interacts with a vibrational mode of the particle and emits at a different wavelength (Raman scattering). The wavelength emitted could be longer (s tokes) or shorter (anti S tokes). In elastic light scattering, the interraction between the incident electromagnetic radiation field and the particle are governed by the Maxwell equations in a medium. Mie in 1908 provided with an exact solution for the interaction of a electromagnetic field a nd a single homogeneous sphere [59] In F igure 6 2 a sperical particle is represented and a coordinate system is defined according to the direction of propagation of the electromagnetic radiation and of its polarization Therefore the direction of the scattered

PAGE 118

118 Figure 62 Spherical coordinate system in elastic light scattering Given a monochromatic incident radiation of intensity I0 and intensity of scattered radiation vertically or horizontal polarized with respect to the scattering plan could be expressed by for each scattering angle : 2 2 01 22sin 4 IIi r (0 61) 2 2 01 22cos, 4 IIi r (0 62) where r is the distance from the particle, and i1 and i2 two intensity functions. On the larger size of particle s Mie scattering converges to the geometric optics law s, and for small particles it agrees with the theory develop ed by Rayleigh in 1871 [60] This simplifies Mies complex intensity function into simple r expressions under the assumption of nonabsorbing molecules. The R ayleigh theory is based on the assumption that the particle see s a unifor m external electric field, and that this field penetrates the particle faster than the time period of the incident electromagnetic wave. These two assumptions may be illustrated by the expressions below:

PAGE 119

119 2 1, a (0 63) 1, m (0 64) w here m is the complex index of refraction of the particle, and a its radius. This theory provides the scattering and absorption total cross section expressed as: 2 22 6 221 32scatm C m (0 65) 22 3 21 Im, 2absm C m (0 66) Using the equations above, one can discuss the dependency of the collected signal intensity over the size of the particle and the laser w avelength. The scattering cross-section varies with the 6th while the absorption cross section varies only with the 3rd power. Given that the extinction cross section is the sum of these two cross sections, one can see that the absorption contribution wi ll be predominant for small particles over the scattering term However the differential extinction parameter will scale with the 6th power of particle size so that the scattered signal of large aromatic structure will completely overwhelmed that of a diat omic molecule. This technique gives a very strong signal for soot particle from the inception regime to the surface growth regime, corresponding to size s from 10 to several hundreds of nanometer s The soot formation mechanism will be discussed in more deta il below The scattering -cross section of soot being several orders of magnitude higher than those of nitrogen, a very strong scattered signal could be emitted. However for a poly disperse system of

PAGE 120

120 particles, the soot volume fraction could be obtained onl y if one assume s the value of a complex index of refraction. Scattering signals being highly dependent on both the size particle (i.e. cross section) and the number density, the interpretation of the signal could lead to high uncertaint ies considering the wide range of diameter s covered from the early soot precursor s to the large particle s (from nanometers to hundred of nanometers). While accurate soot number density and volume fraction measurements could not be obtained without precise calibration and sampling, this technique is very interesting to detect the transition from a non-sooting regime to a sooting regime as the back scattered intensity emitted will increase suddenly or the transmitted signal will drop Another technique has been developed in the last decade to provide soot volume fraction s : the laser induced incandescence (LII ) A high energy laser beam will cause the soot particle temperature to rise to about 4000K and then the blackbody radiation from this soot is observed. Several papers showed that the emitted intensity is proportional to the soot volume fraction. Approaches to calibrat e the LII signal using other techniques have been developed for soot volume fraction measurement s Elastic light scattering was used in combination with LI I in the RAYLIX technique developed by Geitlinger et al [61] .In addition, on -line extinction calibration has been created by Axelsson using two beams from a Nd Y AG laser: one at the fundamental frequency (1064nm) for the LII shot and a less energetic at the double frequency (532nm) for extinction measurement s [17] These two types of calibration suffer also from the same dependen ce on the soot refractive index Wander Wal used gravimetric sampling cali bration [62] thereby avoiding the need for refractive index In this technique, the

PAGE 121

121 soot particles created in the flame w ere mixed in a gravimetric chimney using a swirling flow of nitrogen in order to get spatial homogeneity before sampling. Knowing the gas flow and the density of the sample the soot volume fr action can be estimated to calibrate the LI I Signal. This approach could be of interest for the current system, where a homogeneous spatial distribution in soot particle could be expected due to the high level of mixing in the combustion zone so that ther e is no need for the gravimet ric chimney used by Wander Wal. Moreover, time-resolved LI I has a main strength as being able to provide the soot particle sizes by measuring the decay time of the emitted ra diation. Another advantage of LII is that it shows a rapid rise in LII signal with laser fluence and then a plateau. If one operate s in this plateau by using sufficient fluence of laser the laser beam absorption will not affect the emitted signal However, v aporization effects could occur when the soot is r aised to the vaporization temperature under large laser fluence s as investigated by Yoder and co -workers. [63] They used LI I in line with light scattering to investigate the vaporization effect and did not notice vaporization effect s for a excitation las er fluence under 0.1mJ/cm2. Finally, the line of sight optical access is very suitable to absorption measurements The absorption spectroscopy consists of tuning a laser beam in order to match one or several electronic/vibrational transitions of the prob ed molecules. A spectrally resolved signal is collected and compared to a beam which does not travel into the combustion zone. The analysis of the spectral absorption profile can provide different information depending on experimental setup. One can obtain the number density of species such as O2, H2O, CO, CO2, NO, NO2, OH, NH3, HF, H2S, and CH4, or the temperature and the velocity of the flow through the Doppler shi f t and profile

PAGE 122

122 broadening [64] Diode lasers are convenient light sources for absorption measurements since they are inexpensive for the applications concerned in combustion and can be tuned based on the injection current. Most of the molecules of interest have transitions in the infra red around 1.55 m [64], so that the grating optics should be set for this domain. Finally, tunable diode lasers offer the ability to tune on and off the absorption line, allowing real -time compensation for laser transmission and therefore accurate, real -time species profile measurements. Stability studies Once the stability of n heptane has been characterized by reaching stability limits over different loading s the current system can be used to investigate the intrinsic stability performance of exotic mixtures Two different type s of investigation are of inter est f or this system. First, the use in low h ydrogen content syngas flames in gas turbine combustor s could be prevented by its poor stability performance. The flexibility of the current system enables the variation of two parameters to study the stability perfor mance of syngas First, the residence time could be varied with the air flow rate so that different reactant concentrations and temperatures affect the oxidation chemistry. Then, the hydrogencarbon monoxide ratio could be varied on the flight so that stab ility limits could also be reached on this dimension. Another investigation is motivated by the challenge of curbing emissions abatement in gas turbine combustors. The technology called flameless combustion regime or m ild combustion offer s a promising opportunity to curve simultaneously NOx, CO, CO2 and soot. The flameless r egime is characterized by a non -luminous regime a homogeneous temperature and distributed reaction zone. This regime is obtained with a

PAGE 123

123 high inlet temperature and a low adiabatic flame temperature as a result of an oxygen rarefaction due to dilution. In practical systems, this is achieved by means of internal or external recirculation. The transition to the flameless regime on a semi closed cycle system has been observed with a sharp de crease in luminosity [31] This could be interpreted by a lower soot yield of the flame. The i nvestigation of the high temperature vitiated regime could be made through the stability limits characterization for different dilution gases. Indeed, divergences have been observed between experimental and numerical studies concerning the enhancement of the operability limits with an increase in the exhaust gas recirculation between Singh et al. [6 5 ] and Marek [66] The investigation could be performed by fixing t he flow of air and dilution gases, and by varying the mass flow of fuel before reaching blowout. The present system offers two dimensions to reach a regime in which significant soot particles are formed in order to study the dilution effect on the soot formation process. First, in a premixed regime, the fuel to air rat io can be adjusted in order to reach the soot inception regime of the fuel generally for equivalence ratios above 1.9. The second dimension concerns the degree of premixing of the fuel and ai r mixture. Indeed the temperature of the liquid fuel injected into the air stream could be adjusted to decrease the amount of pre vaporization. In this case, a local flamelet regime could be achieved in the reaction zone increasing the soot formation proce ss. Soot characterization The drop of luminosity at the transition with the flameless regime may be investigated with the current system through soot characterization over different regime s As mentioned earlier, optical techniques to obtain soot volume fraction, size and number density are very sensitive to the size of the molecule and particles probed.

PAGE 124

124 Th erefore, prior to implementing one or an optical scheme an insight in the soot particle formation is necessary. Indeed, we can formulate the question: Could soot precursors be generate d in a controlled manner on this system ? What type of molecule would be present at the current optical access? Soot particles are generated in gas turbine combustor under certain conditions. They are often undesirable a s they represent an environmental and health hazard in the one hand, and decrease the combustion efficiency in the other hand, due to an enhanced radiation and as a byproduct with respect to water and carbon dioxide. The mechanism which rules the early formation of this particle and their further oxidation is very complex and includes gas phase chemistry, surface gas chemistry and particle agglomeration. They include a wide range of particles size from a nanometer to hundreds of nanometer. The early stage s of formation involve the formation of aromatic rings from various fuel radicals such as acetylene and propargyl radicals to form benzene or phenyl. Benzene leads to higher polycyclic aromatic hydrocarbons (PAH) by the H -abstraction, acetylene addition de scribed by Frenklach. [67 ] .The surface of the PAH molecule will grow by this way until it reaches the inception regime and then coagulation and agglomeration between particle s takes place. Finally under solid-gas phase chemistry the soot particle is oxidized by O atoms, OH radical and O2. Different stages in this mechanism have been studied using different means. Here, the current work will focus on the premixed flame configurations. Early gas chemistry with few aromatic rings has been studied using fuel pyrolysis in shock tubes coupled with extinction measurements by Alexiou and Williams for b lends of toluene and nheptane [68 ] Pyrolysis involves mixture of fuel and diluents so that there is no fuel oxidation reaction.

PAGE 125

125 Table 61 Selected experimental studies on soot formation. J.S.R: jet stirred reactor, P.F.R plug flow reactor, DMA: d ifferential m obility a nalyzer, T.E.M: t ransmission e lectron m icroscopy, G.C: g as chromatograph, EX: e xtinction, M.S: m ass s pectrometer, F.I.D: Flame Ionization Detector amu: a tomic m ass u nit., N.S low angle n eutron scattering. Author ( Particle ) Fuel Device phi Temperature res time size Soot level Exp Manzello 2007 [14] ( Soot inception ) C2H4/Air JSR/PF R 1.9 2.2 15401620K 5 42ms 4 70nm 0 50 mg/ m3 DMA TEM Nenniger 1983 [11] ( Soot and PAH ) C7H8/Air JSR/PF R 1.37 2.14 15601670K 4 6ms 0 140 GC C2H2/Air 0.8 2.94 12801640K Wang et al 2002 [6 9 ] ( PAH ) C2H2/O2/Ar Premixed laminar 0.6 2.5 Ti: 330424K Tm: 1698 1787 0 230ms 5 7nm N: 0 107 cm-3 NS Alexiou 1994 [67] ( PAH and HC) C7H8/C7H16 S hock tube 0 1 % mol Ar 15002400 K 0 0.5ms 0. 7 %mol EX C 7 H 8 /C 8 H 18 Macadam 1997 [ 70 ] ( PAH/C2H2 ) C2H2/Air /CO2/N2 JSR/PF R 2.2 15801650 K 0 30ms 128 398 amu GC MS GC FID Different studies have been done on the soot inception and particle growth using laminar premixed flame s or well -stirred reactor -plug flow reactor s A setup in which a well -stirred reactor is followed by a plug flow reactor has been demonstrated to have many advantages when investiga ting the soot inception regime [ 71 ] Using a toroidal jet stirred reactor terminated with a plug flow reactor, Manzello and co -workers measured soot particle size distribution in rich mixtures of ethylene [14] Probing at different distances from the exit of the WS R provide s with exhaust composition after different residence times and so the diff erent step s of the reaction following the soot precursor generation inside the w ell stirred reactor c an be investigated. Wang and co -workers investigate d PAH formation in premixed laminar flames using low -angle n eutron spectroscopy [ 69 ] In laminar premixed flames the conditions are different from the

PAGE 126

126 ones in a jet stirred reactor as the main heat transfer mechanism is heat conduction. This results in much larger residence times of the chemical process as they occur at lower temperature. The reported PAH particles ranges from 57nm which corresponds to the high sensitivity range for small angle neutron spectroscopy. Indeed, one consider s the Arrhenius form of a reaction rate for an elementary step between two products A and B, one gets the reaction rate: 0[] [][],aE RTdA kABe dt (0 67) where Ea is the activation energy and k0 a constant. Residence times in combustion could be compared only at similar conditions in terms of temperature and reactant concentration. The current implementation of the burner has be en motivated by stability limit characterization, so that it involves very small residence times. In C hapter 3, the kinetic simulation provided with an estimat e of the residence time in the primary reactor of the combustion. The total residence time range could be estimated as going from 0.5 to 4ms for the glo bal reaction zone contained in the l iner. The temperature s will range from 1300 to 2200K. In the actual configuration, one cannot really consider the zone downstream of the primary zone as a plug flow reactor since there is dilution by the cooling air. At high loadings and within stability limits, the residence time of the mixture in the primary zone is low but the concentration of reactants being high, the overall reaction rate will be faster. In this configuration, one can intuitively think that only the early stages of the soot mechanism c ould be investigated, as they concern h ydrocarbons and PAH.

PAGE 127

127 Brown and co -workers modeled the kinetics of ethylene air mixtures in a perfectly stirred reactor for various reactor r esidence times and temperature [7 2 ] Their simulations are based on the soot formation mechanism developed by Frenklach. Based on these results, if one run s mixture s of ethylene and air with rich equivalence ratio from 2 to 3 in the current implementation (0.54ms), one can expect to observe mainly small PAH in the system, wh ere size is on the order of a nanometer to several nanometers. Another study has been carried by Singh based also on a Frenklach mechanism but in which particle sizes are modeled on a different scheme [73 ] Based on his model, for a reactor residence time of 1ms, the particle sizes ranged from 2 to 12nm for a flame temperature between 1500 and 2000K. Particles over 50nm should be observed for residence time between 5 and 10ms. These two considerations have to be taken into account for the choice of the optical device. Futures studies and implementation s : The design of this novel combustion system has been thought to offer certain flexibility for a parametric study and a continuous development of this system. With the actual implementation s the effect on the stability performance of the following different parameter c an be stud ied : 1 Rotation of the injector to change the jet impingement angle and the flow split in the primary zone ( F igure 2-4). 2 Modification of the local equivalence ratio of the injected fue l air mixture. One injector could run under a different composition to create an a symmetry in the flame. 3 Staged combustion with two injectors running rich and oriented toward the rear plate and two injectors running lean and oriented toward the exhaust (f igure 6.4). This sequential combustion scheme could be of interest for stabilizing techniques in aerocombustor s

PAGE 128

128 Figure 63 Staged combustion configuration where the two rich injectors are shown Further geometrical changes could be done with the foll owing modifications : 4 To simplify the first design, no nozzles have been implemented on the injector. The presence of a slight elbow close to the injection, this may lead to a disturbed jet velocity profile. Some inserts can be pressed afterward on the inj ectors to include a converging nozzle in the flow. This can be done after some first experimental observations on the combustion regime. 5 Modification of the inner tube ( Figure 6 -4 A) to act on the size of the primary reaction zone. 6 Machining of blank injector with smaller or larger diameter depending whether the interest is in higher flow rates (stability) or lower flow rates (soot) while still providing a good mixing 7 Implementation of 8 secondary air holes in the primary reaction zone for stage d combusti on implementation (Figure 64 B). 8 Modification of the fuel preparation system to increase the pre-vaporization time with a longer premixing section or to evolve in a diffusion combustion scheme with a smaller tube.

PAGE 129

129 9 Implementation of a tube downstream of the liner so that the primary reaction zone could be continued with a reaction zone close to a plug flow reactor for soot inception regime investigation. Some additional diagnostics to combustion species and temperature can be possible with additional im plementations: 10. Implementation of an additional window as shown in Figure 65 or downstream of the reaction zone to provide access for a 90 scattering. 11. Implementation of a transverse nitrogen guided probe to perform real time sampling toward a GC MS (Gas ChromatographMass spectrometer) or a SMPS (Scanning Mobility Particle Spectrometer). This would enable accurate species analysis (GC MS), and particle size distribution (SMPS). In addition to these plans, the design of an afterburner downstream of this multiple injector combustion chamber would be very beneficial in terms of emitted species. Indeed, at high flow rates close to the stability limits, the combustion is incomplete in the primary reaction zone, and the dilution with the cooling may not completely oxidize the intermediates and unburned hydrocarbons created in the primary zone. A B Figure 64 Different geometric modifications : removal and change of the inner tube A), addition of 8 secondary air holes B). Secondary air

PAGE 130

130 Figure 65 Additional UV win dow port for 90 optical access. Finally, an advanced reactor model is of primary importance to understanding of the combustion behavior of this burner and to implement adequate diagnostics. Reacting turbulent flow simulations are an unavoidable step towar d a more accurate reactor network approach. An analysis has to be performed in order to choose the adequate simulation in terms of computing time, complexity and predictions between different model such as the LES, DNS, or other turbulent reacting flow model coupled with simple chemistry. (See C hapter 3, Kinetic M odeling) Based on this simulation, a much more complex reactor neural network model could be created and used to compare with the reference fuel performances for retrofeeding. The current model us ing 3 reactors and implemented with nheptane

PAGE 131

131 mechanism could be upgraded with an additional soot formation mechanism and a resolved reactor network including a larger number of CSTR. A good matching between the kinetic simulation and the experimental resu lts is a necessary step to provide qualitative data on fuel oxidation kinetics. Conclusion s A novel burner design has been created for combustion research. Based on different simulations, the surrounding combustion system has been design ed in order to ful fill different functions such as cooling the wall of the reaction zone, supplying, metering, and controlling the air and finally mixing, atomizing and vaporizing the liquid fuel. The flexibility criterion in terms of flow rate of air, mixture composition and temperature has been of primary importance during the whole design process. The combustor rig has been machined and its implementation has started. This system is expected to operate at residence times in the primary zone between 0.5 and 4ms, with a reactor temperature from 1300 -2400K and an expected maximum total power of 40kW The feed temperature can be varied between 400 and 700K. This rig is implemented with a wide possibility of mixture composition including air, n heptane, methane, different l evel s of syngas, and under dilution by CO2, water vapor, and nitrogen. A line of sight type optical access has been implemented for species and temperature probing at the exit of the primary zone. This system has been designed to provide a laboratory -scale research tool to investigate the stability limits and combustion efficiency of different types of fuel, as well as characterizing the transition to a flameless regim e

PAGE 132

132 APPENDIX A EQUATIONS OF GAS DYN AMICS MODEL These equations are used in the gas dynamics model presented in C hapter 3. The isentropic coefficient is defined by: ,p vC C where Cp and Cv are the specific heats, at constant volume and constant pressure respectively. Isentropic flow 12(1) 2 1 2 2 0 1 1 2 0121 1 12 1 1 2 1 1 2 1 1 2oA M AM p M p T M T M Fanno flow max 22 2 *2 1/2 *2 1/2 2 *1 1 1111 2 ln 1 ln 1 22 1 2 1 2(1) 11 2(1) 12(1) 1 fL D MM M T TM p pMM M M

PAGE 133

133 1 2 2(1) 0* 012(1) (1) p M pM Rayleigh Flow: 22 22 *2 2 *2 22 0 2 0 22 1 0 0(1) (1) 1 1 1 (1) (1)2(1) 1 12(1) 11 TM TM p pM M M MM T T M p MM p Singularities (from the i ncompressi ble flows theory ) In the equation below K is defined as the pressure drop coefficient in the equation: 2 0201, 2 U PPK where 1 and 2 correspond respectively to the upstream and downstream size and U to the flow mean velocity. Sudden section widening: 1 21 A K A Sudden section restriction: 2 1 21 1cA K CA With Cc from 0.6 to 1, depending on the radius, 1 for an infinite radius Sharp e lbow of angle (Weisbach) : 240.457sin2.047sin 22 K

PAGE 134

134 APPENDIX B MACHINING DRAWING The drawings presented in this Appendix B are not necessarily at the scale described in F igure B 1 The reader should refer to the dimensions noted on the drawing s

PAGE 135

135 Figure B 1. Machining drawings

PAGE 136

136 Figure B 1. Machining drawings (continued)

PAGE 137

137 Figure B 1. Machining drawings (continued)

PAGE 138

138 Figure B 1. Machining drawings (continued)

PAGE 139

139 Figure B 1. Mac hining drawings ( continued)

PAGE 140

140 Figure B 1. Machining drawings ( continued)

PAGE 141

141 Figure B 1. Machining drawings (continued)

PAGE 142

142 Figure B 1. Machining drawings ( continued)

PAGE 143

143 Figure B 1. Machining drawings ( continued)

PAGE 144

144 Figure B 1. Machining drawings ( cont inued )

PAGE 145

145 Figure B 1. Machining drawings ( continued)

PAGE 146

146 Figure B 1. Machining drawings (continued)

PAGE 147

147 Figure B 1. Machining drawings ( continued)

PAGE 148

148 Figure B 1. Machining drawings ( continued)

PAGE 149

149 Figure B 1. Machining drawings ( continued)

PAGE 150

150 LIST OF RE FERENCES 1 Y. Huang and V. Yang, Prog Energy Combust Sci 35 (2009) 293 364. 2 D. M. Kalitan, J.D Mertens, M.W. Crofton, E. L. Petersen, J. Propulsion and Power, 23 (2005) 1291 3 J. Natarajan, T. Lieuwen, J. Seitzman, Combust. Flame 151 (2007) 104 119. 4 G. P S mith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M. Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gard iner, Jr., V. V. Lissianski, Z. Qin http://www.me.berkeley.edu/gri_mech/. 5 Q. Zh ang D. R. Noble T. Lieuwen Trans. ASME 129 (2007) 688694. 6 J. P. Longwell, M. A. Weiss, Indust. Eng. Chem. 47 (1955) 1634 1643. 7 F.J. Wright 12th Symposium (International) on Combustion, the Combustion Institute, Pittsburgh, 1969 pp 867 8 W. S. Blazowsk i, R.B. Edelman, and E. Wong, Summary Technical Progress Report, Exonn Research and Engineering, 1980 9 B. R. Bowman, D. T. Pratt and C. T. Crowe, 16th Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1977, pp 819. 10. D. Bradley, S B. Chin, M. S. Draper and G. Hankinson, 16th Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1977, pp 1571. 11. Nenniger J. E., PhD dissertation, M assachussetts I nstitute of T echnology Cambridge, 1983. 12. J. Zelina, D. R. Ballal, Transactions of the ASME 119 (1997) 62-67. 13. J. Zelina, J.W Blust D.R. Ballal, ASME Paper, GT 47 ( 1996) 43 -58. 14. S.L. Manzello, D.B. Lenhert, A. Yozgatligl, G.W. Mulholland, M.R. Zachariah, W. Tsang, Proc Combust Inst. 31 (2007) 675 -683. 15. T. Lieuwen, V. M cDonell, E.Petersen, D. Santavicca, J Eng. Gas Turb Power 130 (2008) 011506 -1 9 16. M.A. Weiss, R.J. Lang, J.P Longwell, Indust. Eng. Chem. 50 (1958) 257264. 17. B. Axelsson R. Collin, P.E. Bengtsson, Appl. Phys B Lasers and Optics, 72, (2001) 367 372.

PAGE 151

151 18. H. J Curran, P. GafIuri, W. J. Pitz C. K. Westbrook, Combust. Flame 114 (1998) 149-177 19. J. Herzler L. Jerig, P. Roth, Proc Combust I nst 30 (2005) 1147 1153. 20. N. M Marinov, W. J. Pitz, C. K. Westbrook, A. M. Vincitore, M. J. Castaldi, S. M. Senkan, Comb. F lame 144 (1998) 192213. 21. N. M Marinov, W.J Pitz, C.K. Westbrook, M. Hori, N. Matsunaga, Proc Comb Inst. 27 (1998) 389-396 22. A. Demirbas, Energy Sources, Part A: Recovery, Utilization, and Environmental Effects 29 (200 7) 1507 1512 23. Srinivasan, W. Ellis, J. F. Crittenden, W. E. Lear, B. Rotavera, E. L. Petersen, Proceedings of ASME Turbo Expo 2008: Power for Land, Sea and Air, June 9 13, Berlin, Germany, 2008, GT 51447. 24. A.K. Gupta, J Eng Gas Turb Power 126 (2004) 9-19 25. K. Maruta K. Muso, K. Takeda, T. Niioka, Proc Combust Inst 28 (2000) 2117 2123. 26. J.A. Wunning J.G. Wunning, Prog. Energy Combust Sci 2 3 (1997) 81 94 27. P.J. Coelho, Comb Flame, 124 (2001) 503 -518. 28. T. Plessing, N. Peters J.G. Wnning, Twenty Seventh Symposium (Int.) on Combustion, The Combustion Institute, Pittsburgh, 1988, pp 3197 3204. 29. A. Cavaliere, M. de Joannon, Prog Energy Combust. Sc i. 30 (2004) 329366. 30. T.S. Nemec W.E. Lear, ASME Paper, 96GT -434 (1996) 1-7. 31. W. Ellis, W. E. Lear, S.A. Sherif,47th AIAA Aerospace Sciences Meeting Including The New Horizons Forum and Aerospace Exposition, 58 January, Orlando, Florida, 2009, pp 255 26 2. 32. Singh W. E. Lear, Proceedings of ASME Turbo Expo 2009: Power for Land, Sea, and Air, June 08 -12, 2009, Orlando, USA, 2009, GT200960272. 33. J. P. Longwell, E. E. Frost, M. A. Weiss, J. Indust. Eng. Chem. 45 (1953) 1629 1633. 34. R. Ballal, A. H. Lefebvre, J. E ng. Power 101 (1979) 343348. 35. Y.D. Wang, Y. Huang D. McIlveen-Wright, J. McMullan, N. Hewitt, P. Eames S. Rezvani, Fuel Processing Tech 87 (2006) 727 736

PAGE 152

1 52 36. M Harris., Small Engine Group, Florida Turbine Technology, Jupiter, Florida, 2009, personal communication. 37. V. W. Greenhough, A.H. Lefebvre, 6th Symposium (International) on Combustion, Reinhold, New York, 1957, pp858869. 38. A.M. Eaton, L.D. Smoot S.C. Hill, C.N. Eatough, Prog Energy Combust Sci 25 (1999) 387 436 39. M. Falcitelli, S. Pasini L. Tognotti, Computer and Chem. Eng. 26 (2002) 1171 1183. 40. A.B. Lebedev, A.N. Secundov, A.M. Starik, N.S. Titova A.M. Schepin Proc. of the Combust. Inst. 32 (2009) 2941 2947 41. A. H. Lefebvre, M.V. Herbert, Proc. Inst. of Mech. Eng. 174 (1960) 463-473. 42. Reeves, National Gas Turbines Establishment, England, NGTE Memo M285, 1956. 43. M. Fishenden, O.A Saunders, Oxford University Press, New York, 1950. 44. Lefebvre A. H., Gas Turbine Combustion, 2nd edition, Taylor & Francis, 1999, pp 5057. 45. W. Kays, M Crawford, B. Weigand, Convective Heat and Mass Transfer, 4th Edition, McGraw -Hill, 2005, pp 302 305. 46. W. Kays. E. Leung, Intern. J. Heat Mass Transfer, 6 (1963) 537-557. 47. G.A. Greene C.C. Finfrock, T.F. Irvine Jr, Exp Therm Fluid Sci. 22 (2000) 145153. 48. W. Thiele, K.J Kundig, D.W. Mur phy, G. Soloway, B. Duffin SAE Technical Paper 900365 49. S. R. Turns, An introduction to Combustion, 2nd Edition, McGraw -Hill, 2000, pp 441-444. 50. R.P Tye, D.R Salmon, Proceedings of the 27th International Thermal Conductivity Conference and the 15th Internat ional Thermal Expansion Symposium, 2005, pp 372-381. 51. S. Kumar P.J. Paul, H.S. Mukunda, Proc. Combust. Inst. 30 (2005) 2613-2621. 52. Y. Mori, K. Kutagami, Intern. J. Heat Mass Transfer 10 (1967) 18011813. 53. M. Conaire, H. J. Cu rran, J. M. Simmie, W. J. Pitz C. K. Westbrook Intern. J. of Chem. Dynamics, 36 (2004) 603-622

PAGE 153

153 54. T. Madhiyanon, N. Piriyarungroj, S. Soponronnarit Energy Conv. and Manag. 49 (2008) 1202-1210. 55. R.A. Shandross J.P Lon gwell, J.B. Howard, Combust. Flame, 85 ( 1991) 282284. 56. W.E. Kaskan Proc. Combust. Inst. 6 (1957) 134 143. 57. A. C. Eckbreth, Laser Diagnostics for Combustion Temperature and Species, 2nd edition., Combustion Science and Technology Series, vol. 3, Gordon and Breach Publishers, New York, 1996 58. N. Zhang, Z. Tian, Y. Y. Leng, H. T. Wang, F. Q., J.H. Meng, Science in C hina Series D Earth Sciences 50 ( 2007 ) 11711178. 59. G. Mie, Ann. Physik, 25 (1908) 377-445 60. L. Rayleigh, Philos. Mag., 41, (1871) 274279 reprinted in Scientifics papers by Lord Rayleigh, Vol 1, No.8, Dover, New York, 1964. 61. H. Geitlinger, Th. Streibel, R. Suntz, H. Bockhorn, Proceedings of the 27th International Symposium on Combustion, The Combustion Institute, Pittsburgh,1998, pp1613 1621. 62. L Vander W al., Z. Zhou, M. Y. Choi, Combust. Flame 105 (1996) 462-470. 63. G. D. Yoder, P.K. Diwakar D.W. Hahn, Appl. Optics 44 (2005) 42114219. 64. M. G Allen, Meas. Sci. Technol. 9 (1998) 545 562 65. B. Si ngh, W.E. Lear, 7th International Energy Conversion Engineering Conference, AIAA, Denver, Colorado, Aug. 25, 2009, AIAA 2009-4570. 66. C. J. Marek R.R. Tacina National Aeronautics and Space Administration, Springfield, Va, 1976, NASA technical memorandum 3464. 67. M. Y. Frenklach, Phys. Chem. Chem. Phys 4 (2002) 2028 -2037. 68. A. Alexiou, A. Williams, Fuel 74 ( 1995 ) 153 -158. 69. H. Wang B. Zhao, B. Wyslouzil, K. Streletzky, Proc Combust. Inst. 29 ( 2002 ) 27492757 70. S. Macadam, PhD Dissertation, Massachussetts Institute of Technology, Cambridge, 1997. 71. F.W. Lam J.P. Longwell, J.B. Howard, Twenty Third Symposium (International) on Combus tion, The Combustion Institute, Pittsburgh, 1990, pp 1477 -1484.

PAGE 154

154 72. N. J. Brown, K. L. Revzan, M. Frenklach, Twenty Seven Symposium (International) on Combustion The Combustion Institute, Pittsburgh, 1997, pp 15731580. 73. B. Singh, PhD candidate, Department of Aerospace and Mechanical Engineering, University of Florida, Gainesville, 2009, personal communication.

PAGE 155

155 BIOGRAPHICAL SKETCH Benoit Fond was born in Angouleme, France, the fourth of a six -child family. After two years of preparatory program in m athematics and p hysics at the Lycee Montaigne in Bordeaux, he passed the entrance examination of Arts et Metiers Paristech where he pursued the 3 years diplome dingenieur This degree is equivalent to a Master of Science in the French G raduate School s of Engineering also called Grandes ecoles. He studied m echanical and i ndustrial e ngineering from 2006 to 2008 and enrolled at the University of Florida in August 2008 to pursue a Master of S cience in m echanical e ngineering as part of a double diploma.