Spray Diagnostics of Low NOx Air Blast Atomizers at Ambient and Elevated Pressure

Transcription

1 ILASS Americas 28th Annual Conference on Liquid Atomization and Spray Systems, Detroit, MI, May 2016 Spray Diagnostics of Low NOx Air Blast Atomizers at Ambient and Elevated Pressure S. D. Pack *, J. A. Ryon, and G. A. Zink Spray Diagnostics and Research Laboratory United Technologies Aerospace Systems West Des Moines, IA USA Abstract Spray diagnostics measurements of 5 different air blast atomizers at ambient pressure conditions are presented. These atomizers were derived from an atomizer originally developed as part of a combustor design for NASA's Environmentally Responsible Aviation (ERA) N+2 advanced, low NOx combustor technologies program. The main objective of the presented research is to probe the performance of the air blast atomizer designs that could be utilized in this combustion system. The atomizer designs were developed extensively by utilizing CFD with the intent of rapidly and thoroughly mixing fuel and air to minimize NOx emissions. This paper provides comparisons of the air flow field predicted by CFD to that measured by Particle Image Velocimetry (PIV) at ambient air-only test conditions. Additionally, the atomizer s behavior with both air and fuel was investigated using Laser Induced Fluorescence (LIF). In these tests, a laser dye (Pyromethene 597) was added to MIL-PRF-7024 Type II test fluid and excited by a laser sheet to produce the LIF results. These air and liquid PIV/LIF results are presented and compared to air-only PIV results as well as a comparison made with multiphase CFD simulations. These 5 concepts were finally tested in a medium pressure combustion rig at UTRC to obtain emission measurements. This emission data is presented and compared to CFD predictions in this paper. * Corresponding author: spencer.pack@utas.utc.com

2 ILASS Americas 28th Annual Conference on Liquid Atomization and Spray Systems, Detroit, MI, May 2016 Introduction As part of NASA s Environmentally Responsible Aviation (ERA) N+2 initiative, United Technologies Aerospace Systems (UTAS) developed combustion technologies for a low-nox combustor [1]-[2]. This program built on UTAS s history of Multipoint Lean Direct Injection (MLDI) concepts [3] which have been shown to have very low emissions of oxides of Nitrogen (NOx). Recent developments from UTAS [4] include radial staging of injectors, converging combustor geometries, and the use of air-blast fuel injectors in addition to the traditional air-assist fuel injectors which were solely used in previous MLDI concepts. It is the intent of these recent developments to advance the practicality of MLDI combustion technologies by increasing the operating range of the combustor, reducing the number of nozzles, as well as improving low-nox emissions to a goal of 75% reduction from the ICAO standard adopted by CAEP 6 at engine pressure ratios of at least 55. UTAS has worked with NASA and the University of Cincinnati to demonstrate concepts with these capabilities [9-11]. The main objective of the presented research is to probe the performance of the air blast atomizer designs that could be utilized in future MLDI combustion systems. Computational Fluid Dynamics (CFD) was used to develop five innovative atomizer designs intended to rapidly and thoroughly mix fuel and air to minimize NOx emissions. This paper provides comparisons of the air flow field predicted by CFD to that measured by Particle Image Velocimetry (PIV) at ambient air-only test conditions. Additionally, the atomizers behavior with both air and fuel was investigated using Laser Induced Fluorescence (LIF). In these tests, a laser dye (Pyromethene 597) was added to MIL-PRF-7024 Type II test fluid and excited by a laser sheet to produce the LIF results. These air and liquid PIV/LIF results are presented and compared to air-only PIV results. These 5 concepts were finally tested in a medium pressure combustion rig at the United Technologies Research Center (UTRC) to obtain experimental emission measurements. This emission data is presented and compared to CFD predictions. Physical Description Although the previous NASA research program [15] focused on how an entire MLDI combustor array operates, it can be advantageous to focus solely on individual fuel injectors to reduce the uncontrolled variables associated with multi-nozzle interactions. Therefore, in the current work, five different air-blast injectors are examined (see Figures 1-4 below). The presented air-blast concepts are unique over other traditional air-blast atomizers because of the placement the fuel between a high swirl inner air circuit and a nonswirling outer air circuit (Patents Pending US 13/664,785, US 13/665,497, US 13/665,568). Additionally, the percentage of the total air that flows through the center air channel is greater than seen in other traditional air-blast atomizers. The intent of these features is to promote rapid mixing of fuel and air in close proximity to the fuel injector, leading to a compact flame and low residence time to help reduce thermal NOx. The rapid mixing also minimizes any interaction with neighboring fuel injectors in an MLDI array. The representative cross section of Concept 1, Outer Air Blast and Concept 2, Intermediate Air Blast is shown in Figure 1. The only difference between the two concepts (not shown) is the total air effective area of the atomizer tip and the flow splits between the two air swirler circuits. Concept 2 has slightly more air through the atomizer since this air blast atomizer was located closer to the richer burning pilots when arranged in the NASA MLDI combustion design. These concepts are considered the baseline concepts since they are unaltered from the NASA ERA program, and were previously presented at the ILASS 2013 conference [14]. Figure 1. Example cross-section of concepts 1 & 2. Although the baseline nozzles demonstrated good mixing and performed well in NASA testing, UTAS wanted to improve on these designs to achieve lower NOx emissions. The three additional concepts presented here represent UTAS s efforts in these areas. Figure 2 shows an iso-view of one of these designs, Concept 3, The Claw. This concept improved upon Concepts 1 and 2 by increasing the overall size of the injector by 14%. The balance of air between the inner and outer circuits was done via CFD simulations. The radial growth of the nozzle allowed for the coannular air jets to be reduced in radial thickness while maintaining a similar AC D, allowing for increased surface area of the fuel where it is injected between the shearing air jets. In addition to growing radially, the inner air swirl vanes were also redesigned, again using CFD to optimize their shape to reduce unnecessary separations and total pressure losses. Finally, the outer air cap supports were redesigned in an attempt to reduce vane wakes through the outer air circuit. It was discovered previously that the vanes in the outer swirler caused wakes that potentially reduced the effectiveness of rapid mixing [14].

3 Using Concepts 1 and 2 as a guide The Claw moved the vanes further upstream to allow wakes to be minimized at the point of air and fuel interaction. Figure 2. Iso-view concept 3, The Claw. Figure 3 shows Concept 4 termed the Multipoint concept. This concept utilized the baseline air configuration but replace the fuel distribution ring with three individual pressure atomizing tips. These pressure atomizers delivered the fuel at three discrete injection points and capitalized on fuel pressure to create fuel atomization. The goal of this injector was to use the finer droplet size created through pressure atomization in combination with the high shear of the air blast swirlers to promote even more rapid mixing. It was also thought that the fine spray from this nozzle could be sufficient for it to be used as the pilot injector in the MLDI array as well as a possible outer/intermediate injector. This injector could be used to decrease NOx emissions from the pilot injector while still maintaining required ignition characteristics. wake separations, thus improving the overall aerodynamic efficiency. CFD was utilized extensively in this process of developing the air circuits to minimize wakes and maximize this air-fuel shear interaction. Additionally, a traditional slotted prefilming fuel circuit was used to distribute the fuel instead of the fuel preparation used in the NASA injectors. However, this fuel circuit was not developed to the same level as the original injectors, and could likely be improved in future designs. Experimental Setup In this experiment, a laser based stereo PIV measurement technique is applied to the spray field produced from the five MLDI atomizers. A single atomizer is contained within a metal chamber that is pressurized with air. A pressurized liquid line is fed directly into the atomizer. The spray is created inside an open test chamber that is at ambient room temperature and pressure conditions. The incoming flow rates are then varied and the pressure and temperature of the supply air and liquid are measured. Figure 3. Iso-view concept 4, Multipoint. The final concept which was designed and tested, called the Bugle, is shown in Figure 4. This concept improved upon the baseline configurations by including a radial vane inner air swirler (Patent Pending US 13/654,176). This inner air swirler includes a radial convergence throughout the length of the swirl vanes, which are intended to reduce unnecessary total pressure losses and efficiently shape the velocity profile to encourage fuel and air mixing. In addition, the outer air vanes were also refined to reduce the separations associated with these vanes. This was done in part by moving the air vanes into the mid passage of the outer air circuit to decouple the inlet separations from the vane Figure 4. Iso-view concept 5, Bugle. The advanced air-blast atomizer configurations were mounted in an air box that was pressurized at 3%, 4%, and 5% P/P across the atomizer tip. Only the 4% P/P air pressure test points are presented in this paper. The air delta pressure was the difference between the pressure measured via a pressure tap just upstream of the nozzle tip and the ambient pressure in the lab. See Figure 5 for test atomizer air box layout. There were two different setups used for this testing. The first was PIV measurements of scatter laser light off of small seeded particles in the air stream. The second was liquid PIV measurements using laser induced fluorescence (LIF). These two setups will be discussed separately in the following two sub sections. The PIV system used to take the measurements was a complete adaptive PIV system for standard planar Stereo-PIV from LaVision. This system included a 532 nm double-pulsed Nd:YAG laser, nanopiv Litron laser (1200 mj / 4 ns), with optics to produce about a one millimeter thick light sheet on the central axis of the flow. The delay between the two laser pulses depended on the flow conditions that were being tested. As the

4 incoming flow conditions are changed, the velocity of the spray flow changes. Thus, the time delay t, needs to be changed to maintain the accuracy of the PIV measurements. The timing of the system is controlled by a programmable timing unit (PTU9). Two CCD cameras (Imager Pro X 4M) each with a Nikon AF Micro Nikkor 105 mm lens with a double shutter feature is used to capture two images per camera for each sample. The images are then stored on a computer for further processing. LaVision PIV software, DaVis v8.2.1, is used to construct the velocity vector fields. The interrogation window is circular 32 x 32 pixels with a 50% overlap. The velocity vectors are analyzed by the DaVis post-processing program that detects and deletes erroneous vectors; masks shadow regions; computes the vorticity, turbulent kinetic energy, and Reynold's stress; and computes the ensemble averages. Only the velocity vectors and contours are presented in this paper. The ensemble averages presented in this paper is comprised of the average of 500 instantaneous velocity vector fields measured at a ~15Hz sample frequency. Air Inlet (3X) Pressure Tap Flow Straightener Atomizer Face Figure 5. Air box layout for atomizer testing. High pressure combustion rig testing of individual nozzles took place at United Technologies Research Center at the Jet Burner Test Stand (JBTS). The experimental facility at UTRC is a self-contained combustion facility having eight test cells. The first five cells are designed for hot flow or combustion testing. The combustion test cells are each approximately 30x17x18 ft with two test centerlines in each cell. Test cells six through eight are designed for cold flow testing (no combustion). Steady state high-pressure air is supplied to all test cells at flow rates of up to 24 lb/sec at pressures of up to 385 psi. Higher air flow rates of up to 150 lb/sec may be obtained by "blowing down" three 5000 cu ft storage tanks which are part of the system and are pressurized to 400 psi by the compressors. At full pressure, the tanks hold 30,800 lbs of air. The high-pressure air supply is dried to a dew point temperature of -20 F and may be heated to 900 F at an 18 lb/sec air flow rate. In addition, there are two 720 kw electric air heaters available for use with the high pressure air supply. These heaters will provide heated air flows up to 5 lb/sec each at 600 F or 3 lb/sec each at 1200 F. Hydrocarbon fuels are supplied to the test cells at pressures of up to 1,200 psi and flow rates up to 35 gal/min from any one of 3-1,000 gallon, 3-4,000 gallon or 1-10,000 gallon above ground storage tanks located in the nearby fuel storage yard. Hydrogen, oxygen and nitrogen gas systems, at pressures up to 2,400 psi, are also provided. The hydrogen and oxygen are stored in remotely located trailer truck cascades of 230,000 and 65,000 SCF, respectively. The nitrogen is stored in a 15,000 SCF tank, which is charged by a LN vaporizer-compressor system. A natural gas supply system capable of supplying up to 0.35 lb/sec at up to 750 psia is also available. Liquid propane, supplied from a remote 1000 gallon tank is pressurized up to 700 psi. The JBTS is supplied with cooling water from a 6 in. city water main for providing test model cooling. Exhaust flow cooling is provided by a Closed Loop Cooling System that greatly reduces the water consumption of the facility. PIV Air Only Measurements PIV is an optical method of flow visualization used in education and research. It is used to obtain instantaneous velocity measurements and related properties in fluids. The fluid is seeded with tracer particles which, for sufficiently small particles, are assumed to faithfully follow the flow dynamics. In these measurements, only the gas flow from the nozzle is measured by a standard stereo PIV technique. The liquid circuit is turned off, so there are no liquid droplets in the flow. The compressed air supplied to the air box is seeded using Bis(2- ethylhexyl)sebacate (DEHS) via a LaVision Aerosol Generator (PN ). The sizes of these DEHS droplets are less than one micron. The air only measurements used a stereo PIV technique to measure the air flow field. The PIV setup is as mentioned and the CCD cameras utilize a band pass filter (BP , # ) from LaVision that allows the scattered laser light to pass thru while limiting background emissions. The air box was oriented face down for all testing. For axial plane test results, the cameras are setup approximately 90 apart facing opposite sides of the laser sheet. The cameras were set up for forward scatter particle imaging and were ~45 off of the laser sheet. A Scheimpflug mount is utilized to allow the two cameras to focus on the laser sheet through the entire measurement area. This mount uses the Scheimpflug Principle to allow the measurement plane to be in focus even

5 though the cameras are at different angles to the laser measurement plane [5]. 500 image pairs (2 images per camera) were collected for all presented test points at a sample frequency of ~10Hz. For the off axis tests, the cameras were placed upstream of the flow field and the cameras were operated in a back scatter mode. There was approximately 90 between the two cameras with each camera approximately 30 above the measurement plane looking downward at the measurement plan. The measurement plane started at 12.5 mm from the atomizer face and was moved incrementally to 37.5 mm. For this work only the 19 mm plane is presented. The test area is vented/exhausted in the downward direction. This downward velocity was not measured but is expected to be less than.5 m/s. The time delta between laser pulses was 7 µs which corresponded to a droplet displacement of 8 to 10 pixels at peak velocity flow field areas. PIV Liquid Spray Measurements The liquid PIV experiments used a stereo PIV technique to measure the spray flow of MIL-PRF-7024 type II (fuel substitute), mixed with Exciton laser dye Pyromethene 597 (P597). P597 readily absorbs into the fuel substitute [6], [7] and does not require premixing the dye with alcohol or other diluents. The liquid flow rate was tested at several different mass flow rates with 4% air P/P. In this technique, only the velocity of the liquid is measured as the liquid drop-lets form the seeding" particles that are tracked in the PIV algorithm. The laser pulse excites the P597 in the water, and it begins to phosphoresce. The CCD cameras are both filtered to capture P597 s phosphorescent frequency, 582 nm. The lens filter used on each camera was an assembly of two filters from Midwest Optical Systems (band pass filter BP and long pass filter LP550-52). The Mie scattering of the laser light is filtered out of the image and only the phosphorescent light is recorded. Thus, the position of the droplets can be more accurately determined. The flow meter used to measure the fuel substitute is a Max 120 Flow Computer utilizing a positive displacement flow meter model As with air only measurements, the laser pulse time delay varies with the air flow rate. The air box and atomizer again were mounted in the downward direction. The cameras were mounted as discussed in the air PIV section. As a reminder only the 19 mm plane is presented in the work. Experimental Results The first measurements completed and presented in this work are the air only velocity measurements at 4% air P/P at ambient pressure. Figures 6, 8, 10, 12, and 14 show the axial velocity contour plots of each configuration through the atomizer center plane. The atomizer face is within 1 mm of the top of each contour. This allowed reflections to be minimal during this testing. Figures 7, 9, 11, 13, & 15 show the tangential velocity contours at the 19mm plane downstream of the atomizer face, Figures 6 and 7 show the baseline velocity contours of the Outer air blast injector configuration. The peak total velocity is approximately 70 m/s. As expected the peak velocity is near the atomizer face/exit. The higher velocity nonswirling outer air quickly dissipates as it mixes with the center swirling air, enlarging and forming a spinning jet. Figure 7 clearly shows the small wakes that are present from the air circuit. This is apparent from the peak velocity zones that appear, recede, and then reappear around the circumference of the flow field. In addition, the number of peak velocity zones corresponded to the number of vanes in the outer air circuit. Figure 6. Mid plane V tot contours, Outer Figure 7. 19mm plane V tot contours, Outer Figures 8 and 9 show the other baseline velocity contours of the Intermediate injector. As expected the contours are very similar to the Outer nozzle. It has a higher air effective area which may help to wash out some of the wakes seen in the Outer configuration.

6 There is also a slight ovaling in the air flow contour that may be due to some slight irregularities in the atomizer construction. The outer and intermediate concepts both had expected results as seen from previous testing of the same designs [14]. The goal of the present work was to improve upon this design by enhancing mixing and decrease combustion emissions. Figures 10 and 11 both show the first optimized concept termed The Claw. As seen, the velocity peaks (~60 m/s) are slightly lower than that of the outer configuration. The wakes are also considerably reduced. There are three high velocity zones whose cause is not readily apparent. There tends to be a slight spreading or widening of the air flow field as the air travels downstream. It is unknown at this time what effect this might have on nozzle to nozzle interaction in the MLDI combustor array. Also, the recirculation zone down the center of the atomizer was larger than that of the baseline concepts. This was a desired feature and would be explored regarding its effect on emissions. Figure 10. Mid plane V tot contours, The Claw Figure 8. Mid plane V tot contours, Intermediate Figure 9. 19mm plane V tot contours, Intermediate Figure mm plane V tot contours, The Claw Figure 12 and 13 show the velocity contours of the Multipoint injector concept. The velocity peaks tend to be slight lower than the baseline concepts. However, the measured flow field seems to demonstrate larger, more irregular wakes than that of the outer and intermediate configurations. These wakes are even seen in Figure 12 in a pattern reminiscent of a barber pole (high and low velocity regions that spin downstream). Since the multipoint design s air circuits were identical to those used on the outer and intermediate nozzles except for a slight obstruction downstream of the inner air vanes, it was expected that the wakes would also be similar to the baseline case. This doesn t seem to be the case. The three internal pressure atomizers required slightly more physical space than the fuel circuit on the baseline atomizer which required slight modifications to the air circuits that may be responsible for these differences. The recirculation zone down the center is smaller than The Claw configuration and very similar to the baseline concepts.

7 Figure 12. Mid plane V tot contours, Multipoint Figure mm plane V tot contours, Multipoint The final configuration tested termed Bugle is shown in Figures 14 and 15. This configuration shows the highest max velocities of all the configurations (~75 m/s). This is expected since it was optimized via CFD to have reduced total pressure losses within the air circuits in order to deliver the highest velocities at the airfuel interface. The higher velocity is even seen to extend further downstream as noted by the higher velocities at the 19 mm plane. It is interesting to note that the velocity profile seems to be the most uniform of all configurations. There are four higher velocity zones but the peaks seem to be closer in value to the velocity minimums. This was the CFD goal for this atomizer to have the least amount of wakes and most uniform velocity profile. The next phase of injector testing was the liquid droplet PIV to determine the velocity flow field of the liquid droplets. This velocity flow field, combined with the laser induced fluorescence images is then used to create a mass flux contour of each atomizer. The mass flux images are useful in understanding the distribution of the fuel downstream of the nozzle face. Figure 14. Mid plane V tot contours, Bugle Figure 16 thru 19 show the liquid velocity and mass flux contours of the baseline configurations at 4% air P/P with 11.3 kg/hour of fuel (~50 psi fuel pressure). Measurements were taken at multiple distances from the nozzle face, but the 19 mm plane is the only plane presented in this work. As seen the flow field of the liquid fuel is significantly different than that of the air only configuration. The velocity max is ~30 m/s compared to the air flow rate of ~50 m/s. This is expected since the air velocity is what is used to atomize the fuel in a pure air blast atomizer. The fuel leaves the atomizer at a relatively slow rate compare to that of the air and is accelerated by the air stream. The interesting thing to note is the decreased recirculation zone down the center of the flow field. The rotation of the flow field is still counter clockwise (air and fuel are corotating) but it appears that there is not a recirculation zone down the center. This is confirmed on both baseline configurations. A key feature of the baseline injectors is clearly seen in the mass flux contour. The mass flux contours appear to be very uniform. There is a slight heavy side on each injector on different sides but very good distribution from three flow slots. The fuel distribution passages are a patented feature of these two injectors and were an area open to redesign on the other configurations in an attempt to improve emissions. Figures 20 and 21 are the liquid velocity and mass flux contours of The Claw configuration. As seen the liquid velocity peak is higher (~40 m/s) than that of the baseline. This is somewhat surprising since the air velocity of The Claw is slightly lower than that of the baseline configurations. This could be due to the fuel having a slightly higher velocity at the atomizer exit. However, this also would be unexpected, as the flow number was targeted to be the same as the baseline injectors, and thus the fuel velocities should be similar.

8 circulation at the center and there is only a small amount fuel present in this area. The slight ovaling seen in the air velocity contour is also seen in the liquid mass flux. It would be expected that this feature would recur since the air is a major contributor in liquid distribution and mixing. Figure mm plane V tot contours, Bugle Figure mm plane liquid V tot contours, Interm. Figure mm plane liquid V tot contours, Outer Figure mm plane M f contours, Outer The mass flux of The Claw shows a uniformity improvement over the baseline configurations. The fuel distribution is also slightly wider than the baseline configurations as well. The center low fuel zone of the mass flux contour is larger than the baseline zone. This is expected since the air contours showed this larger zone. The liquid velocity contours show very little re- Figure mm plane M f contours, Intermediate The liquid flow field and mass flux contours of the multipoint injector are not presented in this work. The atomizer was damaged after combustion testing resulting in unreliable spray results. Thus, the results are unavailable at this time. Figures 22 and 23 show the liquid velocity and mass flux contours of the Bugle concept. This concept shows the fuel velocity to be higher than the baseline which is expected since the air velocity was higher in this concept. The liquid velocity peaks tends to be in a slightly tighter circle but this trend is not seen in the mass flux plot. The uniformity of this concept seems to be the least uniform of the configurations. It is important to note that the mass flux measurements were completed after combustion testing. There is a possibility that this configuration could have suffered a small amount of contamination in one of the fuel passages,

9 which may explain the mass flux non-uniformity. The fuel circuit used for this nozzle was also not as fully developed as the other nozzles tested and could result in this measured nonuniformity. m/s, with the mid plane peak at 75 m/s and the 19 mm plane peak at 75 m/s. Figure mm plane liquid V tot contours, Claw Figure mm plane liquid V tot contours, Bugle Figure mm plane M f contours, The Claw Computational Analysis and Results CFD analysis of the aerodynamic flow fields produced by each of the 5 air blast nozzles was also performed to corroborate the experimental results. The first set of tests for comparison with atmospheric PIV used OpenFOAM CFD code to perform the computations. The method described in [13] was employed which included using a DDES turbulence model. Timeaccurate (2 nd order discretization in space and time), compressible, single-phase (air-only) analysis was performed using atmospheric conditions similar to the lab measurements. The boundary condition specified for all CFD cases is for the 4% P/P air only conditions. The results of the CFD air-only analysis are shown for each nozzle at the Mid-Plane and at 19mm plane similar to those shown for the experimental results (see figures 24 through 33). The contours are shown on the same contour scale as the PIV results. The units are Figure mm plane M f contours, Bugle Generally, the comparison with experimental results shows a good similarity between prediction and measurement. Some discrepancies are noticeable but are likely attributed to small-scale differences between the modeled geometry and actual test hardware. Several features of the geometry, even when machined to a very fine tolerance, have been shown to cause large changes within the flow field. Additionally, the mesh quality that was used in all of the analysis is considered a coarse industrial mesh and therefore may sensitize the results. One additional potential source of error is the location of the slice planes, as a small difference in the position of this slice (even within 1 mm) can make a large difference in the resulting flowfield contours. Considering these sources of error, the comparisons seen here are considered favorable.

10 to show the most significant amount of wakes (it was not optimized by CFD in the early design stages). Based on these results, it is felt that the multipoint configuration has significant room for design improvement. Figure 24. Mid plane V tot contour predictions, Outer Figure mm plane V tot contour predictions, Intermediate Air Blast Figure mm plane V tot contour predictions, Outer Air Blast Figure 28. Mid plane V tot contour predictions, The Claw Air Blast Figure 26. Mid plane V tot contour predictions, Intermediate Air Blast It is interesting to note that the CFD velocity predictions are not free of wakes. A significant CFD effort was completed at the concept design phase to minimize these wakes. As seen, CFD tended to over predict the size of the wakes but it was extremely useful in generating concepts that minimized these wakes. The Claw design tended to show the least amount of wake effects of all the configurations. The multipoint design seemed Figure mm plane V tot contour predictions, The Claw Air Blast

11 significance here. The intent was to decrease temperature hot spots and create a more uniform temperature distribution. This should result in lower levels of thermal NOx emissions, which is the dominant source of NOx in these regimes. As expected the multipoint configuration had the highest temperature profile which in turn lead to the highest emissions predictions and measurements. Figure 30. Mid plane V tot contour predictions, Multipoint Configuration Figure mm plane V tot contour predictions, Multipoint Configuration Figure mm plane V tot contour predictions, Bugle Air Blast CFD simulations were run for each of the 5 injectors at 1 rig test point and 2 separate FAR conditions. For these combusting CFD runs, FLUENT (ANSYS) CFD code was used and the methods described in [4] were used. Specifically, a steady-state (2 nd order spatial discretization) solver was used with the k-ω SST turbulence model. Fuel entered the system through the use of the Discrete Phase Model (DPM) to simulate the liquid sprays and combustion was modeled using the Non- Premixed Mixture-Fraction PDF. Finally, the NOx postprocessor from FLUENT was used to integrate the NOx production. The boundary conditions matched the rig test point and were taken at T3 = 900 degrees F, P3 = 250 psi, and P/P = 4%. The two FARs used were FAR=0.035 and (which correspond to φ=0.514, φ=0.441 respectively). Figure 32. Mid plane V tot contours predictions, Bugle Additionally, combustion CFD was used to predict the amount of NOx emissions generated at a rig test of the nozzles. Figures 34 through 38 show the combustion CFD results for the five different nozzle configurations showing the temperature profile in the top image, velocity magnitude in the lower left image, and the discrete phase concentration in the lower right image. The temperature distribution predictions are of the most Figure 34. Combustion CFD results Outer Air Blast

12 The multipoint injector relies on pressure atomization instead of prefilming liquid fuel circuit to atomize and distribute the flow, and is found to be more sensitive to input boundary conditions of the fuel. The prefilming air-blast nozzle style used in the other 4 nozzles have low momentum fuel delivery around a prefilming annulus so the boundary conditions for those injectors had less sensitivity to the fuel boundary conditions. Additionally, the turbulence model and steady-state conditions are potential sources of error within the simulations. Figure 36. Combustion CFD results Claw Air Blast Figure 35. Combustion CFD results intermediate Experimental Results Each injector was tested at UTRC at conditions ranging from 77 to 250 psi P3 and 500 to 1000 F T3. These covered the range from idle to scaled takeoff as tested at NASA Glenn Research Center under the ERA program. As such, they provided a direct comparison to data from that prior test. The original NASA test successfully demonstrated the viability of the high shear air blast nozzles in reducing NOx, so it made for a good baseline condition to see any improvement with the new atomizers. Figure 39 shows experimental data, as well as the CFD predictions, for one set of test points, at 900 F, 250 psi inlet air. With NOx reduction as the primary thrust of this research, that data will be focused on. Each injector was tested at multiple different fuel-to-air ratios, and emissions indices (EI) of NOx, CO, and unburnt hydrocarbons (UHC) reported. When it appeared that each injector was approaching blowout conditions, as determined by a marked rise in hydrocarbons, the test was moved to another condition. This allowed the test to avoid blowout and restart. While this may have resulted in minor loss of data, the range of fuel-to-air ratios where the nozzles were expected to operate was covered with good fidelity for each nozzle. Figure 37. Combustion CFD results multipoint Figure 38. Combustion CFD results Bugle Air Blast Most immediately, one notices that the CFD prediction for the multipoint is quite different from the test data, and that the multipoint emissions are higher than any of the other nozzles. As mentioned previously, there was uncertainty in modeling the pressure atomizer spray of the multipoint nozzle, so this could explain the difference between predicted and recorded results. Additionally, it is likely the spray penetrates too far into the combustor, as seen in the CFD simulation, causing extended hot spots relative to the other designs, which causes the higher NOx.

13 Figure 39. NOx Emissions predicted by CFD (circles) compared to emissions data (x s). Looking closer, the NOx levels of the outer baseline nozzle increase at high fuel-to-air ratios more rapidly than the other three designs which appear to lie more closely to one another. Of those three, the trends are the same, with the Bugle concept highest, then the intermediate baseline, and the Claw concept lowest NOx. As stated previously, the original designs performed well at NASA, so it is no surprise that the single injector also sees low NOx levels throughout the run. However, the proposed changes in diameter, inner air swirler, and outer air wake minimization have shown benefits in reducing NOx further. At the cruise fuel-toair ratio of 0.030, this is an approximately 15% reduction in EINOx. The CFD comparison is favorable, although there is some error, and the predictions were unable to determine the rankings between the three as to which had the highest and lowest NOx. That said, the ability of the combustion CFD to predict values close to those reported can be a useful tool once more calibration is done on the analysis process and other potential sources of error are reduced. Conclusion The atomizer designs presented in this paper were developed extensively by utilizing CFD with the intent of rapidly and thoroughly mixing fuel and air to minimize NOx predictions. The current research confirms that the actual atomizers created behave as predicted. Although not an exact match, the provided comparisons of the air flow field predictions by CFD matched fairly well to those measured by PIV at ambient air-only test conditions. Additionally, the atomizers behavior with both air and fuel was investigated using LIF. For these purposes, a laser dye (Pyromethene 597) was added to MIL-PRF-7024 Type II test fluid to produce the LIF results presented in this work. The air and liquid PIV/LIF presented were compared to air-only PIV results at similar pressure drops and significant differences were noted in the flow fields due to the added momentum of liquid. Mass flux measurements calculated for the fuel by performing liquid PIV measurements and utilizing the LIF signal were also presented. These mass flux measurements were not compared to mechanical patternation results, which is planned for future work. CFD NOx emission were also presented and compared to actual combustion tests performed at UTRC. These results show favorable comparisons and demonstrated the ability of these air blast injectors. Nomenclature AC D Effective Flow Area CAEP Committee on Aviation Environmental Protection CFD Computational Fluid Dynamics DDES Delayed Detached Eddy Simulation P/P Percent pressure drop taken across the nozzle EI Emissions Index (grams of constituent/kilogram of fuel) EINOx Emissions Index of NOx ERA Environmentally Responsible Aircraft FAR Fuel to Air Ratio ICAO International Civil Aviation Organization MDLI Multipoint Direct Lean Injection N+2 Technology available for next generation or for the 2020 time-frame NOx Oxides of Nitrogen OPR Overall Pressure Ratio P3 Compressor Discharge Pressure (Nozzle Inlet Pressure) T3 Compressor Discharge Tempearture (Nozzle Inlet Temperature) UTAS United Technologies Aerospace Systems References 1. Collier, F., Overview of NASA s Environmentally Responsible Aviation (ERA) Project, 48 th AIAA Aerospace Sciences Meeting (2010). 2. Suder, K., Delaat, J., Hughes, C., Arend, D., Celestina, M., NASA Environmentally Responsible Aviation Project s Propulsion Technology Phase I Overview and Highlights of Accomplishments, 51 st AIAA Aerospace Sciences Meeting, Grapevine TX (2013) 3. Tacina, R. Mao, C., Wey, C., Experimental Investigation of a Multiplex Fuel Injector Module With Discrete Jet Swirlers for Low Emission Combustors, AIAA (2004) 4. Prociw, L., Ryon, J., Goeke, J., Low NOx Combustion Concepts in Support of the NASA Environmentally Responsible Aircraft Program, Proceedings of the ASME Turbo Expo 2012, Copenhagen, Denmark (2012) 5. Walker S., Two-axes Scheimpflug Focusing for Particle Image Velocimetry, Measurement Science and Technology, 2002 Vol. 13.

14 6. Dvorak D., Pack, S., and Hu H., An Experimental Investigation on Spray Flows Exhausted from a Coswirling Air-blast Nozzle, AIAA, June 5, Saeki S, and Hart D P, Investigation on YAG(532) Laser Dyes for Oil Film Thickness and Temperature Measurement" Proceedings of PSFVIP-3, Maui, Hawaii, March 18-21, Kreyszig E., Advanced Engineering Mathmatics Seventh Edition, John Wiley & Sons ISBN , Dolan, B., Rodrigo, V., Munday, D., Gutmark, E., Zink, G., Pack, S., Goeke, J., OH* Chemiluminescence in a Multipoint Combustion System: Steady State and Limit Cycle Behavior, 51 st AIAA Aerospace Sciences Meeting, Villalva, R., Dolan, B., Munday, D., Gutmark, E., Zink, G., Pack, S., Goeke, J., Medium Pressure Emissions of a MultiPoint Low NOx Combustion System, 51 st AIAA Aerospace Sciences Meeting, Villalva, R., Dolan, B., Munday, D., Gutmark, E., Zink, Ryon, J., G., Pack, S., Goeke, J., Medium Pressure Emissions of a Multi-Point LOW NOx Combustion System, ASME Turbo Expo, Berg T., Deppe J., Schucht T., Voges H., Mass Flux Imaging in Sprays, ICLASS, K.R. Reddy, J.A. Ryon, P.A. Durbin, A DDES model with a Smagorinsky-type eddy viscosity formulation and log-layer mismatch correction, International Journal of Heat and Fluid Flow, Volume 50, December 2014, Pages , ISSN X 14. Pack, S., Ryon, J., Zink, G., Dvorak, D., and Goeke, J., Spray Diagnostics of a Low NOx Air Blast Atomizer for NASA ERA N+2 Program, ILASS Americas, 25th Annual Conference on Liquid Atomization and Spray Systems, Pittsburgh, PA, Zink, G., Ryon, J., and Pack, S., Intermediate Pressure Combustion Research of a Multipoint Low NOx Combustion System, 50 th AI- AA/ASME/SAE/ASEE Joint Propulsion Conference, Cleveland, OH, 2014.

15