2006 UTSR PROJECT. CFD Analysis of Air-Fuel Premixer Design for Gas Turbine Injector of A Rich-Catalytic Lean-burn RCL Combustor

Transcription

1 UTSR Industrial Internship Program 06/04/ /25/ UTSR PROJECT CFD Analysis of Air-Fuel Premixer Design for Gas Turbine Injector of A Rich-Catalytic Lean-burn RCL Combustor MSU Intern: David Thiepxuan Cao UTSR Program Manager: Dr. William Day MSU Academic Advisor: Professor Abraham Engeda Precision Combustion, Inc. Industrial Mentors: Dr. Shahrokh Etemad Dr. Benjamin Baird Mr. Sandeep Alavandi Prepared for: Precision Combustion, Inc. - Gas Turbine Group 410 Sackett Point Road North Haven, CT & South Carolina Institute for Energy Studies College Avenue Clemson University Clemson, South Carolina January 8, 2007 Michigan State University

2 CFD Analysis in Air-Fuel Premixer Design for Gas Turbine Injector of A Rich-Catalytic Lean-burn RCL Combustion Pilot David Thiepxuan Cao ( , caoman2k@hotmail.com) Ph.D. Student, 2006 UTSR Fellowship Michigan State University MSU Academic Advisor: Professor Abraham Engeda ( , engeda@egr.msu.edu) PCI Industrial Mentors: Dr. Shahrokh Etemad ( Ext. 217, setemad@precision-combustion.com) UTSR Program Manager: Dr. William Day ( , billday3@comcast.net) Industrial Site: Precision Combustion, Inc. 410 Sackett Point Road North Haven, CT Objectives: Implemented CFD Analyses in designs of air-fuel premixer and postmixer, pilot cartridge without and with swirl effect by using different turbulent fluid modeling. Achievements: o Finished CFD studies of turbulent fluid modeling for air-fuel premixer design. o Supported RCL tests. o Accomplished CFD studies in designs of air-fuel premixer and postmixer, pilot cartridge without and with swirl effect by different turbulent fluid modeling. Spin-Offs: o Initiated the use of turbulent fluid modeling in Standard K-Epsilon (SKE) model, Re- Normalization Group K-Epsilon Model (RNG KE), and Reynolds Stress Model (RSM) for validating and optimizing air-fuel premixer design of gas turbine injector application. o Concentrated on further details of CFD studies in support of validating feasible and optimal designs for a Rich-Catalytic Lean-burn RCL Combustion. o Recommended possible designs of air-fuel premixer to reduce the unmixedness percentage and provide lower NOx emissions. Comments about the UTSR Fellowship Program: o I improved the air-fuel premixer flow field understanding through CFD modeling. I feel that my understanding of CFD Modeling has significantly increased in the 12-week period. o My industrial mentors were willing and constructive to help and teach me throughout the program and establish a good working relationship. o All the employees at PCI were glad to share the working knowledge, discuss problems and solutions, and assist my work in any way possible. o The pay allowances by UTSR were decent and adequate. A direct deposit system would have been a big help for students who do not have their bank in the state they are working in for the UTSR fellowship. o After my week UTSR Internship, I would definitely recommend these opportunities given by UTSR fellowship to any students who wish to have an opportunity to apply the school knowledge into the industry projects. i

3 Table of Contents Abstract Introduction The Unmixedness Parameter 3 3. Comparison of Descriptions, Usage, Characteristics of Standard K-Epsilon (SKE) Model, Re-Normalization Group K-Epsilon Model (RNG KE), and Reynolds Stress Model (RSM) 4 4. Operating Conditions, Boundary Conditions, Mesh Quality, Solution Controls and Residual Monitors of Air-Fuel Premixer Model for SKE Solution, RNG KE Solution, and RSM Solution Methodology of simulating the air-fuel premixer for the non-reacting flow 7 6. Calculations from Standard K-Epsilon (SKE) Model Solution for the non-reacting flow Graphical Results and Discussion of Standard K-Epsilon Model (SKE), RNG K-Epsilon Model (RNG KE), Reynolds Stress Model (RSM) for the non-reacting flow Conclusion and Recommendation References ii

4 Abstract A Rich-Catalytic Lean-burn (RCL ) combustion system has been previously developed by PCI for operation on natural gas to achieve ultra-low NOx emission with low acoustic turndown. Most significantly, fuel-rich operation from RCL combustion system limits the extent of catalyst-stage reaction based on available oxygen, regardless of the fuel s intrinsic reactivity on the catalyst. Therefore, similar catalyst and reactor performance can be obtained for widely varying fuel types. In addition, catalytic pre-reaction extends the combustor s lean flammability limit for all fuels. This allows the low-temperature combustion for both conventional and lowheating-value fuels with low NOx emissions. In order to support the design of the RCL combustor, this UTSR report presents the CFD analysis in air-fuel premixer design for gas turbine injector of a Rich-Catalytic Lean-burn RCL combustion. The results of air-fuel premixer are presented from three different solution controls of Standard K-Epsilon Model (SKE), Re-Normalization Group K-Epsilon (RNG KE), and Reynolds Stress Model (RSM). The results are as: (1) Contours of Mass fraction of fuel (CH4) at the outlet; (2) Contours of Mass fraction of fuel (CH4); (3) Contours of Velocity; (4) Velocity vector; (5) Contours of Static Temperature; (6) Contours of Turbulent Kinetic Energy; (7) The Unmixedness calculated by Mass fraction of Fuel (CH4) and Mass fraction of Air. This UTSR report presents the results and discussion of Standard K-Epsilon (SKE), Re-Normalization Group K- Epsilon (RNG K-Epsilon), Reynolds Stress Model (RSM) for the non-reacting flow. In conclusion, these results will be used to provide a good foundation in designing air-fuel premixer of a RCL combustor and select the appropriate turbulence model for the next CFD study of improving the Unmixedness calculated by Mass fraction of Fuel (CH4). Furthermore, the Re-Normalization Group K-Epsilon (RNG KE), and Reynolds Stress Model (RSM) probably provide a better result for the CFD analysis for the design of airfuel premixer of RCL combustion. The best simulation model may be the Reynolds Stress Model (RSM), but this model will require more computational time. 1

5 1. Introduction The current gas turbines operating on natural gas offer the lowest NOx emissions without having exhaust-gas after-treatment. This has been achieved by the use of lean-premixed combustion systems. This gas turbine allows NOx emissions near 5-9 ppm. In order to lower emissions, a Rich Catalytic Lean Burn (RCL ) combustion system was developed to even lower emissions (< 2 ppm NOx). The schematics of RCL system are shown in Figure 1.1 and Figure 1.2. In the schematic, the combustion air stream from a gas turbine compressor is split into two parts upstream of the catalyst reactor. The combustion air flow is mixed with all of the fuel in a section called air-fuel pre-mixer. This first mixing stage creates a fuel-rich mixture. All the fuel-rich mixture contacting and passing over the catalyst reactor has insufficient oxygen to completely oxidize all of the fuel. Therefore, the lack of oxygen inside the catalytic reactor limits the extent of catalyst-stage reaction and permits the temperature to be below the instantaneous auto-ignition temperature. This effectively prevents the flashback damage. The combustion air flow is also used to cool down the catalyst reactor. The air stream remains free of fuel and this precludes the flashback or the auto-ignition to the cooling stream. Catalyst Cooling Combustion Air Burned Gas Fuel Premixer Catalytic Reactor Post- Catalyst Mixing Figure 1.1. Schematic of Fuel Rich-Catalytic Lean-burn (RCL ) system At the exit of the catalytic reactor, the catalyzed rich-fuel/air stream and the cooling air are rapidly mixed to produce a fuel-lean, reactive mixture in a section called air-fuel post-mixer (post-catalyst mixing). This second mixing stage creates a fuel-lean mixture and occurs prior to the final combustion in the main combustor of the gas turbine. The auto-ignition delay enables near-perfect mixing to be achieved before the start of the homogeneous combustion. This facilitates the downstream combustion to happen in a lean-premixed mode with ultra-low NOx emissions. The ignition and the combustion of the fuel-lean reactive mixture, which is achieved in the downstream combustion section, produce the flame. This is called Rich Catalytic/Lean-burn, or RCL combustion. This approach avoids both soot formation and high temperatures, which in non-catalytic RQL (Richburn/Quench/Lean-burn) designs lead to high NOx formation. In the RCL system, fuel-rich reactions occur at moderate temperatures on the catalyst surface. For the two-stage catalytic combustion system tested (Figure 1.2), the measurements of performance are the catalyst temperatures in the first stage (catalyst reactor) and the combustor emissions from the second stage. 2

6 Air Exhaust ~ Compressor Catalyst Cooling Combustion Turbine Generator Fuel Premixer Catalytic Reactor Post-Catalyst Mixing Figure 1.2. Schematic of Fuel Rich-Catalytic Lean-burn (RCL ) in Lean combustion system 2. The Unmixedness Parameter The unmixedness parameter is defined as the standard deviation of the local fuel-air ratios (over an assumed discretized domain) divided by the overall fuel-air ratios (the average of the fuel-air ratios). Under the lean conditions, the high percentage of the unmixedness of the reactants (fuel and air) can cause large heat release rate oscillations and the combustion instability. In this report, the unmixedness of fuel-air pre-mixer is based on mass fraction of fuel and air at outlet plane and defined as: Standard_deviation Unmixednes s = Average The design requirement of the injector (the fuel-air premixer) is that the fuel-air pre-mixer produces a uniform fuel-rich mixture ratio. The unmixedness parameter is used to evaluate the mixing state for fuel-air mixture from the optimal design of the fuel-air premixer. Throughout the combustion chamber, there could be local regions of non-optimum oxygen-fuel that result in gradients in temperature and variations in the mole fractions of the products. This effect, which is created by the higher percentage of unmixedness and persists through the combustion chamber, can produce high emissions, particularly for NOx emission. Furthermore, there are small-scale in-homogeneities in the fuel-air mixture, and the burning rate is not perfectly constant and influenced by variations in the local pressure along the length of the combustion chamber [9]. It is necessary to have a lower percentage of unmixedness for fuel-air mixture after leaving the fuel-air pre-mixer and before entering the catalytic reactor. 3

7 3. Comparison of Descriptions, Usage, Characteristics of Standard K-Epsilon (SKE) Model, RNG K- Epsilon (RNG KE), and Reynolds Stress Model (RSM) a) Standard K-Epsilon (SKE) Model [5] The SKE used the two-transport-equation model to solve the RANS (time-averaged Reynolds Averaged Navier- Stokes) model. The SKE uses the isotropic eddy viscosity assumption. The SKE is robust, widely used, economical, and reasonably accurate. However, it performs poorly for the complex flow involving severe p, separation, strong streamline curvature. The SKE is suitable for initial iterations, initial screening of alternative design, and parametric studies. b) Re-Normalization Group K-Epsilon Model (RNG KE) [5] The RNG K-Epsilon equation improved the ability to model highly strained flows and predict swirling and low Re flows. The RNG K-Epsilon is suitable for complex shear flows involving rapid strain, moderate swirl, vortices, and locally transitional flows such as boundary layer separation, massive separation and vortexshedding behind bluff bodies. The RNG K-Epsilon is good for moderately complex behavior like jet impingement, separating flows, swirling flows, and secondary flows. However, the RNG K-Epsilon has limitations due to isotropic eddy viscosity assumption. c) Reynolds Stress Model (RSM) [5] The RSM used the seven-transport-equation model to solve the RANS model. Reynolds stresses are solved directly with transport equations avoiding isotropic viscosity assumption of other models. The RSM avoids the isotropic eddy viscosity assumption. The RSM is suitable for complex 3D flows with strong streamline curvature and severe pressure gradients, highly swirling flows and rotation like curved duct, rotating flow passages, swirl combustors with very large inlet swirl, and rotation. The RSM is a physically most complete model because it is tightly coupled momentum and turbulence equations. The history, transport, and anisotropy of turbulent stresses are all accounted for. However, it is tougher to converge due to close coupling of equations and the RSM requires more memory, computational effort and time. 4. Operating Conditions, Boundary Conditions, Mesh Quality, Solution Controls and Residual Monitors of Air-Fuel Premixer Model for SKE Solution, RNG KE Solution, and RSM Solution a) Operating Conditions The INLET operating pressure condition was at 4 atmospheres. The INLET operating conditions of fuel CH4 were fuel velocity ( ft/sec), fuel temperature (25 C), fuel mass flow rate ( Lbs/sec). The INLET operating conditions of air were air velocity ( ft/sec), air temperature (578 C), air mass flow rate ( Lbs/sec). b) Boundary Conditions Because of the symmetry of the air-fuel premixer model and the reason of saving of the computational time, the air-fuel premixer model was divided into 14 sub-models. The boundary conditions of air-fuel premixer model were defined as follows: - One wall boundary condition (Wall, Outer wall for Air Tube) - Two wall boundary conditions (Wall, Inner wall for Fuel Tube) - Two wall boundary conditions (Wall, Middle wall for mixture Tube) - One wall boundary condition (Wall, Front wall between the outer Air Tube and the inner Fuel Tube) - Three air-hole interior boundary conditions (Interior, for three Holes) - One air-gap Interior boundary condition (Interior, Air inlet through the gap) - One air inlet boundary condition (Velocity Inlet) - One fuel inlet boundary condition (Velocity Inlet) - One mixture outlet boundary condition (Outflow) - Ten symmetry boundary conditions (Symmetry) 4

8 Figure 4.1. Description of Boundary Conditions in GAMBIT c) Mesh Quality From GAMBIT, the mesh quality of the air-fuel premixer model was examined to guarantee the convergence criteria and the solution reliability. There are three volumes of the air-fuel premixer model, which were meshed using Tetrahedral scheme and size of The air-fuel pre-mixer model had elements. After the meshed volumes of the model were completed, the elements of the air-fuel premixer model had a maximum quality value of and a minimum quality value of Therefore, this model met the 0.98 EQUIANGLE SKEW required. 5

9 d) Solution Controls and Residual Monitors d1) Solution Controls and Residual Monitors of SKE The four equations are Flow, Turbulence, CH4, and Energy. The Under-relaxation Factors are defined as Pressure (0.3), Density (0.8), Body Forces (1), Momentum (0.2, this coefficient is so low so that the convergence solution can be reached with a smaller number of iterations. In the future solution, the momentum coefficient should be 0.4 to 0.8 and this will required more computational time and more iterations), Turbulent Kinetic Energy (0.6), Turbulent Dissipation Rate (0.6), Turbulent Viscosity (0.6), CH4 (1), Energy (0.6). The eight Residuals were as Continuity (10-3 ), x-velocity (10-3 ), y-velocity (10-3 ), z-velocity (10-3 ), Energy (10-6 ), k (10-3 ), Epsilon (10-3 ), CH4 (10-3 ). The Pressure-Velocity Coupling is SIMPLE. The discretizations are Pressure (Standard), Momentum (First Order Upwind), Turbulence Kinetic Energy (First Order Upwind), Turbulence Dissipation Rate (First Order Upwind), CH4 (First Order Upwind), Energy (First Order Upwind). d2) Solution Controls and Residual Monitors of RNG KE The four equations are Flow, Turbulence, CH4, and Energy. The Under-relaxation Factors are defined as Pressure (0.3), Density (1), Body Forces (1), Momentum (0.7), Turbulent Kinetic Energy (0.8), Turbulent Dissipation Rate (0.8), Turbulent Viscosity (1), CH4 (1), Energy (1). The eight Residuals were as Continuity (10-3 ), x-velocity (10-3 ), y-velocity (10-3 ), z-velocity (10-3 ), Energy (10-6 ), k (10-3 ), Epsilon (10-3 ), CH4 (10-3 ). The Pressure-Velocity Coupling is SIMPLE. The discretizations are Pressure (Standard), Momentum (First Order Upwind), Turbulence Kinetic Energy (First Order Upwind), Turbulence Dissipation Rate (First Order Upwind), CH4 (First Order Upwind), Energy (First Order Upwind). d3) Solution Controls and Residual Monitors of RSM The four equations are Flow, Turbulence, Reynolds Stresses, CH4, and Energy. The Under-relaxation Factors are defined as Pressure (0.3), Density (1), Body Forces (1), Momentum (0.7), Turbulent Kinetic Energy (0.8), Turbulent Dissipation Rate (0.8), Turbulent Viscosity (1), Reynolds Stress (0.5), CH4 (1), Energy (1). The eight Residuals of the solution were as Continuity (10-3 ), x-velocity (10-3 ), y-velocity (10-3 ), z-velocity (10-3 ), Energy (10-6 ), k (10-3 ), Epsilon (10-3 ), uu-stress (10-3 ), vv-stress (10-3 ), ww-stress (10-3 ), uv-stress (10-3 ), vw-stress (10-3 ), uw-stress (10-3 ), CH4 (10-3 ). The Pressure-Velocity Coupling is SIMPLE. The discretizations are Pressure (Standard), Momentum (First Order Upwind), Turbulence Kinetic Energy (First Order Upwind), Turbulence Dissipation Rate (First Order Upwind), Reynolds Stresses (First Order Upwind), CH4 (First Order Upwind), Energy (First Order Upwind). 6

10 5. Methodology of simulating the Air-Fuel premixer for the non-reacting flow a) Outline of simulating the air-fuel premixer for the non-reacting flow The pre-mixer model was simulated without having the Volumetric Reaction. The option of Volumetric Reaction was turned off and there is NO COMBUSTION. The solution was run until the solution reached the Convergence. b) Methodology of simulating the air-fuel pre-mixer for the non-reacting flow After the Standard K-Epsilon solution of non-reacting flow had been converged with 2551 Iterations, the Standard K-Epsilon (SKE) solution of non-reacting flow was modified by RNG-K-Epsilon solution of nonreacting flow. The RNG-K-Epsilon solution of non-reacting flow was again converged with 452 Iterations. Next, the RNG-K-Epsilon solution was modified by RSM of non-reacting flow. The RSM solution of nonreacting flow was again converged with 1545 Iterations. 7

11 6. Calculations from Standard K-Epsilon (SKE) Solution (2 Equations), (Non-reacting flow) Calculation 1: INLET mass flow rate of CH4 fuel (from SKE), (Non-reacting flow) The INLET fuel mass flow rate of fuel from experiment is *10-5 kgs/sec and the INLET fuel mass flow rate by surface integrals from Fluent *10-5 kgs/sec. The percentage difference of fuel mass flow rate between the experiment and FLUENT is 100*((5.4221*10-5 ) - (5.2324*10-5 ))/ (5.4221*10-5 ) = 3.85 %. The percentage difference is very small and the solution effectively meets the convergence criteria. This small percentage difference occurs because: (1) Model doesn t include the wall thickness; (2) Model uses the inlet velocity. Calculation 2: INLET Mass Flow Rate of air (from SKE), (Non-reacting flow) The INLET air mass flow rate of fuel from experiment is *10-4 kgs/sec and the INLET air mass flow rate by surface integrals from Fluent *10-4 kgs/sec. The percentage difference of air mass flow rate between the experiment and FLUENT is 100*((2.4091*10-4 ) - ( *10-4 ))/ (2.4091*10-4 ) = 1.7 %. The percentage difference is very small and the solution effectively meets the convergence criteria. Calculation 3: OUTLET Mole Fraction of CH4 (from SKE), (Non-reacting flow) The equivalence ratio from the experiment is Φ = 3.5 (Fuel-Rich Condition) The equivalence ratio, Φ, is defined as follows: ( A/ F ) Stoichiometric MWAir Φ = ; ( A/ F ) 4.76 ( / ) Stoichiometric = a A F MWFuel y 4 MWAir = ; MWFuel = MWCH = 16 ; CH4 = CxHy ; a = x + = 1+ = MW Air ( A / F ) 4.76 Stoichiometric = a = = MWFuel 16 ( ) ( A/ F ) Stoichiometric A/ F = = = Φ 3.5 MassAir NAir MWAir NAir MWFuel 16 ( A/F) = = = ( A/F) * = * = Mass N MW N MW Fuel Fuel Fuel Fuel NAir = 2.72NFuel The mole fraction of fuel CH4 from experiment is N Fuel N Fuel x = x CH = = N + N 2.72N + N 1 = 3.72 Fuel = 4 Air Fuel Fuel Fuel Air The mole fraction of fuel CH4 from experiment is x Fuel = x CH = and the mole fraction of fuel CH4 by 4 area-weighted average surface integrals from FLUENT is x Fuel = x CH = The percentage difference of 4 the mole fraction between the experiment and FLUENT is 100*( )/0.278 = 3.27%. The percentage difference is very small and the solution is reasonable. Calculation 4: Conservation of Mass (Continuity Equation) by Using Mass Flow Rates of INLET Air, INLET Fuel and OUTLET Mixture (from SKE), (Non-reacting flow) The total mass flow rates of INLET air, INLET fuel from experiment is m & = m& + m& = 2.37* *10-5 = 2.9*10-5 kg/s INLET AIR-INLET FUEL-INLET

12 The mass flow rates of OUTLET mixture by surface integrals from FLUENT is m& = 2.9*10-5 kg/s OUTLET The percentage difference of the mole fraction between the experiment and FLUENT is very small. Therefore, the solution effectively meets the convergence criteria ( m & = m& ). INLET OUTLET Calculation 5: Calculation of the Unmixedness at outlet plane (from SKE and Non-reacting flow) The calculation of the Unmixedness, which is based on mass fraction of fuel (CH4) and mass fraction of air at outlet plane of reactor, is as follows: Standard _ deviation Unmixedness = Average Table 6.1. The Unmixedness of mass fraction of CH4 fuel (Y CH4 ) and mass fraction of air (Y AIR ) at outlet plane (from SKE solution) The Unmixedness for CH4 Fuel and Air from SKE solution Mass fraction of CH4 fuel (Y CH4 ) Mass fraction of air (Y AIR ) Standard Deviation Average Unmixedness Percentage 29% 37% In conclusion, the same method of the above calculations can be used for RNG K-Epsilon solution (RNG KE) and Reynolds Stress Model solution (RSM). 9

13 7. Graphical Results and Discussion of Standard K-Epsilon (SKE), RNG K-Epsilon, Reynolds Stress Model (RSM) for the non-reacting flow 10

14 Figure 7.1. Contour of Mass fraction of CH4 at exit by SKE (Above), RNG KE (Middle), and RSM (Below) (non-reacting flow) - The several contours from SKE solution do not completely account for the complex structure of the flow. - The contours from RNG KE solution show the complicated flow patterns. From the contour, one area being close to the wall of the mixture tube shows that the flow has a moderate recirculation. Therefore, the mass fraction of fuel at the local area of the wall of the mixture tube from RNG KE solution is almost zero. - The contours from RSM solution show a highly recirculating flow. 11

15 Fuel Inlet Around the center of the outlet of the mixture tube, the concentration of the mass fraction of Fuel from SKE solution is very low. However, this SKE solution is very inaccurate because this SKE solution does not account for the complex flow. Air Inlet Mixture Outlet Around the center of the outlet of the mixture tube, the concentration of the mass fraction of Fuel from RNG KE solution is higher than the RSM solution. 12

16 Around the center of the outlet of the mixture tube, the concentration of the mass fraction of Fuel from RSM solution is lower than the RNG KE solution. Figure 7.2. Contour of Mass fraction of Fuel (CH4) by SKE (Above), RNG KE (Middle), and RSM (Below) (non-reacting flow) The contours of mass fraction of fuel (CH4) by RNG KE and RSM show more mass fraction of fuel at the downstream than the contours of mass fraction of fuel (CH4) by SKE. The contours of mass fraction of fuel (CH4) by RSM show that the mixing state of RSM solution at the outlet is more realistic than the RNG KE solution. At the downstream, the Fuel mass Fraction around the center of the mixture tube from RSM solution is less than the RNG KE solution. Therefore, the unmixedness percentage from RSM solution is lower than the RNG KE solution. When anisotropy of turbulence (with different properties in different directions) significantly affects the mean flow, the RSM will be considered and also provide the strong coupling between Reynolds stresses and the mean flow. The steady and small variation of velocity from the end wall of the mixture tube in RSM solution allows a good mixing state from highly recirculation flow. Therefore, the RSM solution provides a lower unmixedness percentage calculated by mass fraction of fuel (CH4). 13

17 The momentum mixing state finished before reaching the outlet. The high velocity contour is not around the center of the premixer model at the outlet. The mixing state extends a little further downstream, but the mixing state does not reach the outlet. The high velocity contour is around the center of the premixer model. 14

18 The mixing state extends until the outlet. The high velocity contour is around the center of the premixer model at the outlet. Figure 7.3. Contour of Velocity profile by SKE (Above), RNG KE (Middle), and RSM (Below) (non-reacting flow) The velocity contour of air-fuel mixture from RSM shows that the mixing state extends until the outlet. On the contrary, the velocity contour of mixture of air and fuel from SKE and RNG KE shows that the mixing state does not extends until the outlet. The max velocity from RNG KE and RSM is higher than the max velocity from SKE. Fuel is very rich at the 11 Spatial Points around the center of the Tube from RSM solution. The RSM solution provides a logical velocity contours. 15

19 The max velocity from RNG KE solution is higher than the max velocity from SKE solution. 16

20 The max velocity from RSM solution is higher than the max velocity from SKE solution. Figure 7.4. Velocity vector by SKE (Above), RNG KE (Middle), and RSM (Below) Flow Path (non-reacting flow) The figure shows the upstream and downstream flow field of the air-fuel premixer. The max velocity from RNG KE and Reynolds Stress Model (RSM) is higher than the max velocity from Standard K-Epsilon Model (SKE). The flow paths of velocity vector from SKE, RNG KE, and RSM seem reasonable. There is no re-circulation from the upstream of the air-fuel premixer. However, the max velocity from RNG KE and RSM solution is higher than the max velocity from SKE solution. This makes the fuel content to be extended until the outlet of the mixture tube. The air flow has the recirculation after the air flow enters three holes and the gap between the outer wall (air tube) and the middle wall (mixture tube). In RNG KE and RSM solution, the air stream flows towards and is reflected from the middle wall. Therefore, the max velocity from RNG KE solution is higher than the velocity from RSM solution. 17

21 18

22 Figure 7.5. Contour of Temperature profile by SKE (Above), RNG KE (Middle), and RSM (Below) (nonreacting flow) In general, the static temperature profiles of the non-reacting flow from SKE, RNG KE and RSM solutions are reasonable. 19

23 20

24 Figure 7.6. Contour of Turbulent Kinetic Energy by SKE (Above), RNG KE (Middle), and RSM (Below) (nonreacting flow) The turbulent kinetic energy around the 3 interior holes from SKE solution is higher at the beginning of the mixture tube (middle wall). The higher turbulent kinetic energy intensity around three interior holes of the mixture tube from SKE probably allows the mixing state to be finished before reaching the outlet of mixture tube. The maximum turbulent kinetic energy intensity is from SKE. The turbulent kinetic energy intensity from the beginning of mixture tube is higher from SKE model. Therefore, the high velocity contour from SKE does not occur around the center of the mixture-tube outlet. The turbulence increases along horizontal walls whereas it decreases along vertical walls. This confirms that the solutions were reasonable. There is more turbulent kinetic energy at the beginning of the mixture tube by RNG KE. The maximum turbulent kinetic energy by RNG KE is higher than the maximum turbulent kinetic energy by RSM. The maximum turbulent kinetic energy intensities are almost the same for both RNG KE and RSM. However, the variation of turbulent kinetic energy intensity from the beginning of mixture tube is higher from RNG KE solution than RSM solution. The higher variation of turbulent kinetic energy intensity of the beginning of mixture tube from RNG KE probably allows the mixing state to be finished before reaching the outlet of mixture tube. Therefore, the high velocity contour from RNG KE does not occur around the center of the outlet of mixture tube. 21

25 The results of the unmixedness percentage are tabulated as follows: Table 7.1. Unmixedness from Standard K-Epsilon (SKE) Solution with Non-Reacting Flow Mass fraction of fuel (CH4) Mass fraction of air Standard Deviation Average Unmixedness Percentage 29% 37% Table 7.2. Unmixedness from RNG K-Epsilon Solution with Non-Reacting Flow Mass fraction of fuel (CH4) Mass fraction of air Standard Deviation Average Unmixedness Percentage 76% 16% Table 7.3. Unmixedness from RSM Solution with Non-Reacting Flow Mass fraction of fuel (CH4) Mass fraction of air Standard Deviation Average Unmixedness Percentage 69% 14% Having compared the unmixedness based on mass fraction of fuel (CH4) from SKE solution, RNG K-Epsilon solution, and RSM solution, the results of RNG K-Epsilon solution and RSM solution are very close. The result of SKE solution is completely different with the results of RNG K-Epsilon solution and RSM solution. Therefore, the results of the unmixedness from RNG K-Epsilon solution and RSM solution are more reliable. In addition, when looking at the profile of mass fraction of fuel (CH4) at the outlet, the contour of RSM solution shows the better mixing state than the contour of RNG K-Epsilon solution. In conclusion, the RSM solution probably provides the best option in simulating the premixer model of air-fuel mixture. 22

26 8. Conclusion and Recommendation A high unmixedness percentage of air-fuel mixture in the premixer will be reduced as the fuel-air flows through the reactor. However, any high level of the unmixedness in the downstream flame zone will result in higher NOx emissions and combustion instability. Therefore, the design modification of the current air-fuel premixer, which should have the mixing state occurring right inside the fuel tube, is necessary. This will facilitate the lower level of the unmixedness in the downstream flame zone. From the plot of contours of mass fraction of fuel (CH4), the fuel is unmixed until the downstream. It is important that the plot of contours of mass fraction of fuel (CH4) should not be extended further downstream. The fuel should be mixed at the upstream before reaching the end (the outlet) of the mixture tube. The current design of air-fuel premixer can be added by 6 or 14 small rectangular gaps around the circumference of the fuel tube. This allows the air to get into the fuel tube right at the beginning and the mixing state will start right inside the fuel tube. Therefore, the new modified design could reach the low unmixedness percentage of air-fuel mixture at the outlet of the premixer. Having based on the results of CFD studies of the current air-fuel premixer in a Rich-Catalytic Lean-burn RCL combustor, different designs of the air-fuel premixer should be implemented. These future CFD studies of the modified air-fuel premixer will give a more optimal picture of how to minimize the unmixedness percentage of air-fuel mixture at the outlet. In order to validate this simulation solution, an experiment should be implemented to validate the reliability of the CFD study. From the experiment, the future air-fuel premixer designs may be optimally modified to reduce the unmixedness percentage. A very basic model (SKE) and two more elaborate models of RNG KE and RSM were used to simulate the model. SKE is very dissipative and inaccurate in swirling flows. The RNG K-Epsilon model can yield significant improvements over the standard K-Epsilon (SKE) model for re-circulatory flows. RSM is much more accurate than SKE for swirling (or complex) flows. Since the RSM accounts for the effects of complex flows involving high pressure gradients, strong streamline curvature, swirl, rotation, and rapid changes in strain rate, it has the greater potential to give more accurate predictions for complex flows. On another hand, the geometry of the grid is more affected by the RSM model, and then it is more difficult to find convergent result. One of the advantages of the RSM model is that it is easier to find solutions comparable with experimental results due to the near wall treatment. The SKE models use the simple standard wall function whereas on the RSM allows to choose between different wall treatments. The premixer model had different results because SKE results are not accurate. For this problem, the model should at least use the RNG K-Epsilon model. The effect of swirl on turbulence is included in the RNG K- epsilon model, enhancing accuracy for swirling flows. Results of the RNG K-Epsilon model will be better than SKE, but it is still less accurate than RSM. However, using the RSM model is more computationally expensive. 23

27 In order to improve the CFD analysis of future designs of low-unmixedness-percentage air-fuel premixer, the premixer model may implement the following elements: - Investigate the effects of the mesh size, grid resolution and temporal resolution (First and Second order discretization scheme) by GRID SENSITIVITY STUDY. - In the setup, it is important to better define K and epsilon in the Fuel inlet BC by using the intensity and hydraulic diameter, and this may provide a good solution. - Tetrahedral mesh should be used for a fine mesh on the small inlet holes of the premixer. For convergence and accuracy, a quality mesh of the air-fuel premixer is critical, the EQUIANGLE SKEW should be less than 0.9 everywhere and Moderate aspect ratios should be less than The accuracy of the results from air-fuel premixer requires a second order discretization scheme and smaller residuals. Sometimes, the higher order schemes can create the instability for solutions. When the initial guesses of the solution and the saving of the computational time are desired, the first order scheme and the default values of residuals may be useful for quickly providing the solutions. - Simulate air-fuel premixer by the following option: Simulate a non-reacting flow. (Volumetric option in the species model panel should be turned off) 24

28 Comments about the UTSR Fellowship Program My experience at Precision Combustion, Inc. was the working knowledge of CFD modeling and combustion modeling for validating designs of air-fuel premixer, air-fuel postmixer, pilot cartridge without and with swirl effect. The working atmosphere of the industrial mentor was very encouraging and helpful. My co-workers at Precision Combustion, Inc. were open-minded and available for discussions. This year s UTSR stipend and travel allowance were adequate for the year of 2006 I would highly recommend the UTSR fellowship program to anyone who has an interest in the development and the improvement of gas turbine technology. The week Industrial Internship was a good opportunity so that I could apply the knowledge of Computation Fluid Dynamics (CFD), which I have learnt from Michigan State University. On the one hand, I accumulate the useful knowledge from Precision Combustion, Inc. industrial mentors. Moreover, this internship gave me a chance to practice the teamwork skills and engineering communication skills with Precision Combustion, Inc. industrial mentors and employee. On the other hand, I could invigorate my theoretical and practical knowledge of Computation Fluid Dynamics (CFD). As a result, I feel more selfconfidence in my future work with CFD and Gas Turbine Combustion Technology. I learnt the following lessons from Precision Combustion, Inc. 1. Presentation capability: Summarize the important points; Focus on the improvement of the industrial projects; Provide the productive discussion with Precision Combustion, Inc. industrial mentors. 2. Organizing capability: Plan activities, prioritize and manage resources, attempt for continuous improvement. 3. Decision-making capability: Realize the risk of the methodology; Evaluate the feasible options and the optimized solutions; Exercise the judgment in making the rational decisions for Different Turbulent Fluid Modeling and Combustion Modeling. 4. Analyzing capability: understand complexities; Establish a sound problem and identify the issues. 5. Learning capability: Keep learning continuously by being open-minded and flexible. 6. Communication skills: Articulate the ideas persuasively in both verbal and written form. 7. Networking skills: Build a cooperative partnership for exchange of knowledge, experience and services for mutual benefits. 8. Team skills: Demonstrate commitment and share resources. Acknowledgement I would like to express my deep gratitude and great appreciation to Precision Combustion, Inc. industrial mentors (Dr. Shahrokh Etemad, Dr. Benjamin Baird, and Mr. Sandeep Alavandi), PCI employees and 2006 UTSR financial sponsorship in assisting me to successfully finish my week industrial internship at Precision Combustion, Inc. in North Haven, Connecticut. All the considerate guidance, the constructive discussion, the willing attitude and the valuable advice encouraged and supported me to accomplish the Precision Combustion, Inc. project s goals. 25

29 9. References 1. L.L. Smith, H. Karim, M.J. Castaldi, S. Etemad, W.C. Pfefferle, V.K. Khanna and K.O. Smith (2005). Rich-Catalytic Lean-Burn Combustion for Low-Single-Digit NOx Gas Turbines, J Eng Gas Turbines and Power, 127, pp M. Lyubovsky, L.L. Smith, M. Castaldi, H. Karim, B. Nentwick, S. Etemad, R. Lapierre and W.C. Pfefferle (2003). Catalytic Combustion over Platinum Group Catalysts: Fuel-Lean versus Fuel-Rich Operation, Catalysis Today, 83, pp H. Karim, K. Lyle, S. Etemad, L.L. Smith, W.C. Pfefferle, P. Dutta and K.O. Smith (2003). Advanced Catalytic Pilot for Low NOx Industrial Gas Turbines, J Eng Gas Turbines and Power, 125, pp L.L. Smith, H. Karim, S. Etemad and W.C. Pfefferle (2005). Catalytic Combustion of Gasified Coal for Low-Emissions Gas Turbines, 22 nd Annual International Pittsburgh Coal Conference, September 12-17, FLUENT Inc. FLUENT 5 Users Guide, Vol 1-4, published by FLUENT Inc., Lebanon, New Hampshire, USA (1998). 6. Shaw, C. T. Using Computational Fluid Dynamics, Prentice Hall International (UK), Hertfordshire, UK (1992). 7. Hozef Arif, (1999). Application of Computational Fluid Dynamics (CFD) to the Modeling of Flow in Horizontal Wells, M.S. Thesis, Stanford University. 8. Andrew Campbell Lee, (2003). Experimental Investigation of Liquid Fuel Vaporization and Mixing in Steam and Air, M.S. Thesis, University of Washington Lars Davidson (2003). An Introduction to Turbulence Models. Department of Thermo and Fluid Dynamics, Chalmers University of Technology, Sweden. 11. Tobias Jansson, Benoit Chancenotte (1999). A comparison study between k-e Models and Reynolds Stress Models and a repetition of J.Bredbergs work for a 2D ribbed channel, Chalmers University of Technology, Sweden. 12. CFX-5 Solver Modeling. Turbulence and Near-Wall Modeling. 13. J. G. Janzen, L. B. S. de Souza and, H. E. Schulz, Kinetic Energy in Grid Turbulence: Comparison between Data and Theory, Department of Hydraulics and Sanitation, São Carlos School of Engineering. Journal of the Braz. Soc. of Mech. Sci. & Eng. by ABCM October-December 2003, Vol. XXV, No. 4 / H.K. Versteeg and W. Malalasekera, An Introduction to Computational Fluid Dynamics: The Finite Volume Method, Addison-Wesley, John C. Tannehill, Dale A. Anderson and Richard H. Pletcher, Computational Fluid Mechanics and Heat Transfer, Taylor & Francis,