ASSESSMENT OF FREE STREAM SEEDER PERFORMANCE FOR VELOCIMETRY IN A SCRAMJET COMBUSTOR Jason Howison Advisor: Christoher P. Goyne Deartment of Mechanical and Aerosace Engineering University of Virginia Abstract A free stream seeder designed to roduce articles to conduct article image velocimetry (PIV) in a scramjet combustor was studied. Using a robe and filter aer couled to a vacuum um, samles were collected from the outut of the seeder in conditions near to those that are encountered in PIV studies. The filter aer samles were imaged using a scanning electron microscoe (SEM) to obtain qualitative and quantitative data about the articles being generated. The seeder was able to roduce.27 µm diameter articles, near the target size of.3 µm. The light scattering and flow tracking abilities of the articles were also examined with the results indicating that the articles will erform well in scramjet combustor PIV research. C D C s d P Nomenclature = drag coefficient = scattering cross section = article diameter dv dt = acceleration of the article f c = turbulence frequency I o = laser intensity K s = ratio of secific heats M = relative Mach number P s = total scattered ower R = secific gas constant Re = Reynolds number T = temerature 2 2 u u f = relative fluctuation intensities of article and fluid motions V g = velocity of the fluid V = velocity of the article µ = viscosity of the fluid ρ g = density of the fluid ρ = density of the article = angular frequency of turbulent motion ω c Introduction Particle Image Velocimetry (PIV) has become a common non-intrusive method that is used by exerimentalists to investigate comlex flow fields. The key advantage is that PIV can be alied in high seed flows without the flow being disturbed by the resence of robes 1. However, the technique relies uon using a suitable seeding concentration with aroriately sized, uniform articles to roduce quality results. Proer article size means that the articles are large enough to rovide a strong signal to noise ratio but also small enough to faithfully track the flow. Uniform seeding size hels avoid excessive intensity from larger articles and background noise from smaller articles 2, 3. A number of methods exist for roducing such seeding, but the resent study is concerned with the alication of PIV in the University of Virginia Suersonic Combustion Facility 4,5 where high temerature and velocity make most seeding methods unfeasible. The suersonic combustion facility is a continuous running, oen wind tunnel design that roduces a vitiate-free Mach 2 flow. Mach 5 flight enthaly of a scramjet is simulated using electrical resistance heaters, and suersonic combustion of hydrogen is studied using a single 1-deg unswet ram fuel injector. Temeratures aroach 12 K during mixing and 18 K during combustion. In order to obtain velocity information on the entire flow field resulting from the combustion rocess using PIV, both the fuel and free streams must be seeded with articles. Although PIV data has been taken from the fuel stream 6, no data has been taken from the free stream since an effective seeder has yet to be imlemented. Because the free stream has a flow rate aroximately 25 times that of the fuel stream, a different style article seeder is needed for the free stream. The current work is aimed at the design and assessment of such a free stream seeder and the articles roduced. Extensive literature exists for different methods of article seeding 1-3. Condensation and atomization are the most common for liquid drolets whereas atomization and fluidized beds are the most common for solid articles. Atomizers can diserse solid articles in an evaorating liquid or create high vaor ressure drolets in a low vaor ressure liquid 1. In a fluidized bed, articles are susended in a chamber and then drawn out of the to for disersion into a flow 2. Smoke machines may also be used to seed gas flows. The literature also resents a variety of article tyes. Some of the more common articles used in gas flows Howison 1
include oils, olystyrene, titanium dioxide (TiO 2 ), alumina (Al 2 O 3 ), and silica (SiO 2 ). For high temerature flows, such as the one in consideration that can reach over 18 K, the articles must have a sufficiently high melting or boiling temerature in order to scatter light rior to breaking down. The desired article size is a critical decision in PIV imlementation as well. The choice of otimal diameter for seeding articles is a comromise between an adequate resonse of the articles to changes in the flow, requiring small diameters, and high signal to noise ratio of the scattered light signal, necessitating large diameters 2. Past work at the suersonic combustion scramjet facility has indicated that.3 µm diameter sized articles are a suitable balance between flow tracking and light scattering ability for the current PIV work 6. A free stream seeder using silica articles using a fluidized bed was chosen for imlementation. The seeder design is discussed later. Although the silica articles are secified at.25 µm diameter by the manufacturer, it can not be assumed that the seeders are roducing articles of this size. This is because articles this small tend to agglomerate, esecially after eriods of storage or exosure to humidity. Thus, it is necessary to assess the erformance of the seeding system by measuring the outut article size. This aer resents the design of a free stream seeder and the results of an exeriment aimed at measuring the article outut size by the seeder. In the exeriment, articles were samled from seeded air flow using a robe and filter aer. The article size was determined directly by analyzing images from a scanning electron microscoe (SEM) of the articles on the filter aer. These results will be resented along with a discussion on the theoretical light scattering and flow tracking ability of the articles outut by the seeder. Design The free stream seeder was designed with four main criteria: 1) outut article size and article size distribution, 2) outut article number density, 3) oerating ressure and ressure dro, and 4) controllability. The first oint emhasizes the need to design a seeder that roduces articles near.3 µm in size as required by PIV exeriments. The number density is critical since at least 15 articles are recommended er PIV interrogation region 7. However, the density must not be too high either as this can lead to difficulty in distinguishing individual articles. The seeder also needs to oerate at ressures higher than the flow in which the articles will be injected, but the oerating ressure should be minimized above that oint to reduce ressure vessel design requirements. Likewise, the ressure dro in the seeder should be minimized since a larger ressure dro requires a higher initial ressure. Finally, the seeder should have controllability over the outut number density so that seeding can be adjusted as necessary during PIV exeriments. The free stream seeder is shown schematically in Fig. 2. The flow was slit, with Path A byassing the fluidizer and Path B going to the fluidizer. During the exeriment, all of the flow was directed into the fluidizer with Path A remaining closed. Once articles rose to the to of the fluidizer, they were brought to the exit of the seeder for injection into the flow. Figure 2: Schematic of free stream seeder aaratus. An aroximately 1.5 cm thick layer of stainless steel balls at the bottom of the fluidizing chamber was used in order to diserse random jets being roduced by the orous lug. A article shearing nozzle, which is used in the fuel stream seeder, was not ossible since the flow rates necessary for such a design would introduce too much unheated air into the wind tunnel. The shearing nozzle works by injecting articles transversely into a high seed flow, which effectively shears aart agglomerates. Amorhous silica microsheres with a manufacturer secified diameter of.25 µm +/-.5 µm and article density of 2 kg/m 3 were chosen as seeding articles. Although no melting oint is secified by the manufacturer, amorhous silica s melting and boiling oints are over 17 K and 22 K, which means article breakdown during combustion testing is not exected. The seeder was designed to oerate at 41 kpa with little or no ressure dro, which is well within the Howison 2
design criteria. Additionally, the seeding density could be controlled by adjusting the seeder s suly ressure or by changing the amount of flow between the Path A and Path B. Exerimental Technique Several ossibilities were considered for examining the sizes of articles outut by the seeder. Measuring articles from PIV images taken during exeriments would be the most convenient way of obtaining article size; however, the PIV equiment used at the scramjet combustion facility is only caable of 9 µm resolution, which won t rovide any useful information on the size of sub-micron sized articles. Alternatively, a laser can be used to observe a article s diffraction or aerodynamic flight time 8. The former method correlates the intensity and angle of light scattered by the articles to their size. The latter method looks at the deceleration of a article between two light sources and then determines article size using the known density of the article material. However, since both methods assume sherical articles, incorrect results would be found if the articles are not sherical. This assumtion was not made for the silica articles. In the end, the outut seeded flow was samled directly using a robe and filter aer couled to a vacuum um. The articles could not be samled from the free stream in exerimental conditions due to the high temeratures and velocities involved. Therefore, an exeriment was designed for collecting samles from the seeder that enabled the oeration of the seeder as close as ossible to normal conditions. A flow was seeded that was similar to that of an actual combustion exeriment, and then articles were samled from that flow. The details of the exeriment are resented in Table 1. Table 1: Test Conditions Parameter Free Stream Seeder Gas N 2 Particle tye.25 µm SiO 2 Primary filter aer.1 µm Rotameter flow rate 5 ft 3 /hr Exosure times 3 s, 2 min, 5 min The free stream seeder was tested in a flow through an acrylic tube created by a blower at room temerature. The flow created by the blower was similar to that encountered at the oint of injection in the scramjet combustion facility. An overview of the setu is given in Fig. 4. The article laden air from the seeder was injected against the flow direction using the actual wind tunnel free stream injector. A 1.32 mm diameter samling robe constructed out of Inconel tubing was used and laced aroximately 32 mm downstream of the injector. The velocity of the air flow in the acrylic tube was measured using a itot tube with the results indicating that the flow needed to be samled at.17 m 3 /hr in order to ensure isokinetic samling. Isokinetic samling requires that the flow rate into the robe be the same as the flow rate in the acrylic tube so that biasing towards samling smaller or larger articles does not occur. The air was ulled through.1 µm filter aer using a vacuum um along with a rotameter and vacuum gauge to verify flow conditions. Figure 4: Overview of free stream seeder exerimental setu. Nitrogen was used in the free stream seeder instead of air, which will be used in wind tunnel alications, since comressed nitrogen was readily available. The static ressure of the fluidizer was also changed from normal wind tunnel oerating conditions. Normally, the fluidizer oerates at a maximum of 41 kpa (roughly 16 kpa above the tunnel s total oerating ressure of 25 kpa). Pressures higher than this in the fluidizer will introduce an unwanted amount of unheated air into the wind tunnel free stream. The fluidizer was oerated at 7 kpa above the total ressure in the acrylic tube at 27 kpa. The total ressure in the acrylic tube was 2 kpa. Thus, a higher seeding density could be achieved at higher ressures in the fluidizer during wind tunnel oeration than was done by this exeriment. The samles were imaged using a scanning electron microscoe (SEM). A square, roughly 1 cm in length, was cut out from each samle so that the samles would fit on the SEM mounting blocks and mounted for insection. Samles were also coated with a 1-2 nm layer of gold alladium to reduce image drifting under high magnification due to a charge Howison 3
buildu, which resulted from the filter aer being nonconductive. Image Processing The non-homogeneous nature of the filter aer made it very difficult for comuter rograms to discern between the articles and filter aer. As a result, no automated rocesses were created that could successfully extract quantitative data from the SEM images. Therefore, the image rocessing for the samles was erformed manually. The articles characteristic lengths were measured using an image viewer, where a article s longest linear dimension was defined as its characteristic length, which would also be assumed to reresent its diameter. This length was measured in ixels and then converted to metric units using a reference scale included on the SEM images. The results that will be resented were obtained from the same magnification, 3x. However, the analysis was conducted at other magnifications to check for consistency. Results & Discussion Several of the roerties of the articles from the seeder are resented in this section. First, article shae and size are discussed as obtained from direct analysis of the SEM images. Then the articles abilities to scatter light and accurately track high seed and turbulent flows are examined. Particle Shae Using the SEM to create images of samles also enabled a close u look at the shae of samled articles. Fig. 4 shows articles at 3x. As can be seen in the figure, the rimary silica articles were all nearly sherical. However, agglomerated article sets, seen as one large article in PIV exeriments, were not sherical. Later analysis will assume a sherical shae for all articles and article sets. Particle Size A histogram and cumulative distribution for the article sizes are reorted in Figs. 5-6, resectively. The histogram was truncated at.5 µm since the remaining articles were only a few ercent of the total article count. The seeder s largest measured article was.78 µm and comosed of four agglomerated rimary articles. The data was collected from 227 measured articles in four images. Only the 2 and 5 minute run time samles were used in the analysis since the SEM images from the 3 second run time were blurry and would introduce unnecessary error. Figure 4: SEM image of articles catured on filter aer. The free stream seeder s average outut article size was.27 µm. It is interesting to note that 198 of the 227 articles counted were comosed of only a single rimary article. The average size of the rimary silica articles was.24 µm, which is in agreement with the manufacturer secified size of.25 µm +/-.5 µm. A vast majority of the remaining articles were comosed of only two rimary articles. Thus, it is seen that the seeder is successful at breaking down agglomerates. Percent of Total Particles 3 25 2 15 1 5.2.23.25.28.3.33.35.38.4.43.45.48.5 Light Scattering Particle Characteristic Length, µm Figure 5: Histogram of article sizes. As mentioned earlier, the light scattering ability of a article is critical to its effectiveness at roviding useful knowledge of the flow. Too large of articles will saturate light sensing equiment while too small of articles will only generate noise. The theoretical light scattering cross section was calculated for the articles obtained during the exeriment using the aroach taken in Ref. 9. Howison 4
Percent of Total Particles 1 9 8 7 6 5 4 3 2 1..1.2.3.4.5.6.7.8 Particle Characteristic Length, µm Figure 6: Cumulative article size distribution. Equation (1), given by Ref. 2, gives the relation for light scattering cross section, C s, where P s is the total scattered ower and I o is the laser intensity incident on the article. It was assumed that the articles would scatter light according to Mie theory based on article sizes. It was also assumed that C s scaled to the article characteristic length raised to the sixth ower, and that observation angle could be neglected, which is sensitive to article size. Fig. 7, showing the light scattering cross section, was roduced using the article size distribution shown in Fig. 6 using a linear weighted average. Ps Cs (1) I o The analysis shows that the silica articles scatter light fairly evenly over the range of article sizes. The signal to noise ratio roduced by the articles in PIV tests, therefore, should be excellent for obtaining quality data. The larger articles will not saturate PIV sensing equiment, and the smaller articles will not degrade the data with background noise. Normalized Light Scattering Cross Section 18 16 14 12 1 8 6 4 2.2.23.25.28.3.33.35.38.4.43.45.48.5 Particle Characteristic Length, µm Figure 7: Normalized Mie light scattering cross section. Flow Tracking Flow across an oblique shock wave is first considered in order to investigate each article s flow tracking ability using the aroach used in Ref. 9. Since the ste change in velocity across a shock is nearly instantaneous, studying a article s ability to track such a change is a convenient way to measure its flow tracking ability across to a large disturbance. The case considered is deicted in Fig. 8, where an oblique shock is generated in a Mach 2 flow with a velocity of 13 m/s over a 1-deg wedge. These conditions were chosen as they are tyical of the conditions seen in Ref. 4 and 5. The velocity vectors V 1 and V 2 shown in Fig. 8 corresond to the streamline velocities ustream and downstream of the shock, resectively, and V 1n and V 2n the velocity comonents normal to the shock. Figure 8: Oblique shock formation in air on a 1- degree wedge (reroduced from Ref. 9 with ermission). The aroach taken by Meyers in Ref. 3 was used to calculate the article acceleration after the oblique shock. As exlained in Meyers, since the article density is much greater than that of the fluid, the article motion can be adequately described using only Stoke s drag as the force term. The article acceleration can then be given as follows: dv dt = 3 CDµ Re ( Vg 4 ρ d V 2 ) (2) In Eq. 2, dv /dt is the acceleration of the article, C D is the drag coefficient, defined in Eq. 3a, Re is the effective Reynolds number between the gas and article, defined in Eq. 3d, µ is the viscosity of the fluid, ρ is the density of the article, d is the diameter of the article, V g is the velocity of the fluid, and V is the velocity of the article. In Eq. 3c, M is the relative Mach number, K s is the ratio of secific heats, and T is the temerature. Howison 5
where C D 51.1 CD + M = Re 51.1 1. +.256M CD + M Re 24 Re.82 (3a).28Re C D = +.4 + 1.6e (3b) M Vg V = (3c) K RT s ρ g Vg V d Re = (3d) µ The model resented assumes sherical articles. The inut arameters for the analysis are summarized in Table 2. By alying this article drag model to the article distributions shown earlier, it was ossible to determine the roortion of articles accurately tracking the flow at a distance downstream of the shock. Table 2: Flow tracking calculation inut arameters Parameter Value V n1, m/s 621. V n2, m/s 428.6 Gas density behind shock, kg/m 3.22 Secific heat ratio behind shock 1.34 Gas constant behind shock, J/kg K 287 Static tem. behind shock, K 667 A article was considered to be effectively tracking the flow if its velocity was within 1 % of the fluid velocity. Fig. 9 shows the roortion of articles tracking the flow as a function of the normal distance downstream of the shock. Nearly 9 % of the articles are tracking the flow 2 mm downstream of the shock. However, the diagnostic equiment used in PIV exeriments is caable of resolution on the order of 1 mm where very few articles are tracking the flow. If a strong disturbance, such as a shock, is encountered in PIV studies, this would cause the shock structure to aear smeared in results. A article s resonse to turbulence was also considered. The aroach used by Melling in Ref. 2 is used here. A article was considered to be resonding to turbulence in the flow if the ratio of the fluctuation intensities of the article and fluid motion, the left hand side of Eq. (4a), were equal to.95. The fluid viscosity, µ, in Eq. (4b) was found using Sutherland s model. The angular frequency of turbulent motion, ω c, and the frequency of turbulent motion, f c, are related in Eq. (4c). Percent of Particles Tracking Flow 1 9 8 7 6 5 4 3 2 1 1 2 3 4 5 6 Normal Distance Downstream of Shock, mm Figure 9: Particle tracking behavior downstream of an oblique shock wave. where u u 2 2 f ω = 1 + c C d 2 1 (4a) 18µ C = (4b) ρ ω 2π = c c f (4c) The lot in Fig. 1 shows the ercentage of articles resonding to a given turbulence in the flow. All of the articles are caable of resonding to turbulence frequencies less than 1 3 Hz, and 95 % of the articles will resond to frequencies less than 1 4 Hz. Percent of Particles Resonding 1 9 8 7 6 5 4 3 2 1 1.E+3 1.E+4 1.E+5 1.E+6 f c, Hz Fig. 1: Particle frequency resonse. Conclusion A article seeder designed for use in the free stream of a scramjet combustor has been resented. An Howison 6
exeriment was erformed to assess the outut article size from the seeder. Particles were samled from the seeded gas from the seeder using a robe and filter couled to a vacuum um system. The filter aer was imaged using an SEM so that the article characteristic length of the outut articles could be determined. The seeder roduced an average article size of.27 µm. SEM images showed that the rimary silica articles were nearly sherical. Predictions on the light scattering and flow tracking ability of the articles were made assuming all articles were sherical. Using a Mie scattering model, it was seen that articles over the whole range of sizes obtained from the seeder scattered similar amounts of light. This means that data acquired in future PIV tests using this seeder will have excellent signal to noise ratios. The flow tracking ability of the articles was also resented by using a Stoke s drag model. At 2 mm normal distance downstream of an oblique shock, it was redicted that 9 % of the articles were caable of accurately tracking the flow. Although this is excetional erformance, it does mean that discontinuities in the flow will aear slightly smeared in PIV results. It was also seen that 95 % of the articles are caable of resonding to turbulence frequencies of u to 1 4 Hz. The exerimental results indicate that a fluidized bed can be used with great success in generating nonagglomerated, uniform seeding articles from silica owder for PIV exeriments. Moreover, the articles outut by such a seeder have excellent light scattering and flow tracking roerties. Future work includes imlementing the free stream seeder in the suersonic scramjet combustion facility and evaluating the quality of the data against the redictions resented in this aer. Acknowledgements This work was sonsored by NASA Langley Research Center with Glenn Diskin as technical monitor. The author would also like to thank Kemit Finch, Josh King, and Dan Glanz for their work in designing and imlementing the free stream seeder as well and the Virginia Sace Grant Consortium for their suort. References 1 Raffel, M., Willert, C.E., and Komenhans, J., Particle Image Velocimetry, A Practical Guide, Exerimental Fluid Mechanics, edited by R. J. Adrian, M. Gharib, W. Merzkirch, D. Rockwell, and J. H. Whitelaw, Sringer-Verlag, Berlin, 1998,. 1-59. 2 Melling, A., Tracer articles and seeding for article image velocimetry, Meas. Sci. Technol., 8, 1997,. 146-1416. 3 Meyers, J.F., Generation of Particles and Seeding, Laser Velocimetry, Lecture Series 1991-8, van Karman Institute for Fluid Dynamics, Belgium, June 1991. 4 Krauss, R.H., McDaniel, J.C., Scott, J.E., Whitehurst, R.B., Segal, C., Mahoney, G.T., and Childers, J.M., Unique, clean-air, continuous-flow, high-stagnation-temerature facility for suersonic combustion research, AIAA Paer 88-359, July, 1988. 5 Krauss, R.H., and McDaniel, J.C., A Clean Air Continuous Flow Proulsion Facility, AIAA Paer 92-3912, July 1992. 6 Goyne, C.P., McDaniel, J.C., Krauss, R.H., Day, S.W., Velocity Measurement in a Dual-Mode Suersonic Combustor Using Particle Image Velocimetry, AIAA Paer 21-1761, Aril 21. 7 Keane, R.D., and Adrian, R.J., Otimization of article image velocimeters. Part I: Double ulsed systems, Meas. Sci. Technol., 1, 199,. 122-1215. 8 Iacocca, R.G., German, R.M., A Comarison of Powder Particle Size Measuring Instruments, International Journal of Powder Metallurgy. Vol. 33, Issue 8, 1997,. 35-48. 9 Owens, Z. C., Goyne, C. P., Krauss, R. H., McDaniel, J. C., Assessment of Particle Seeder Performance via Direct Flowfield Samling, AIAA Paer 23-918, January 23. Howison 7