Characterization of the kerosene spray in a swirl-stabilized flame

Transcription

1 Characterization of the kerosene spray in a swirl-stabilized flame Nicolas Fdida 1,*, Christophe Brossard 1, Axel Vincent-Randonnier 1, Daniel Gaffié 1 1: Onera - The French Aerospace Lab, F Palaiseau, France * correspondent author: nicolas.fdida@onera.fr Abstract Combustion processes involved in aeronautic combustors greatly depend on the turbulent flow structure, the characteristics of the liquid fuel spray and its atomization. Simulation of these physical and chemical processes uses complex numerical tools which lean on experimental data. The drop-size of the spray is an important feature which impacts directly on the computation quality. The aim of this study is to obtain a large database on the liquid fuel spray characteristics for a swirl-stabilized flame, under conditions close to industrial combustors. A Phase Doppler Interferometer (PDI) measured the droplet size and velocity distributions for the steady spray flame. Spatial distributions of these data were compared for two operating conditions. A recirculation zone was evidenced from PDI results and the swirl motion of the flame was characterized. The transient ignition phase was investigated, even if the main goal was to study the steady phase of combustion. Results showed similar mean droplet sizes and velocities in both phases. The PDI processes the laser scattering signals from droplets passing through a measurement volume with a high frequency sampling, which is not constant in time. To perform a Fourier transform on droplet time arrival, a resampling processing is required. Such a sampling algorithm, developed at ONERA, was performed for several measurement locations. Frequencies close to 1 khz were evidenced, with values depending on the measurement location in the flame and on the operating conditions. To explain these frequencies, possible causes, such as chamber acoustics or swirl vortex breakdown, are examined. 1. Introduction The combustion process in aeronautic combustors greatly depends on the turbulent flow structure, the characteristics of the liquid fuel spray and its atomization. To obtain meaningful and reliable numerical computations of combustion processes, the physical input parameters of the operating conditions have to be well known. The drop-size distribution of a spray is an important input parameter because it has a direct impact on numerical results quality. The aim of this study is to obtain a large database on the liquid fuel spray characteristics for a swirl-stabilized flame, under conditions close to industrial combustors. For this purpose, a Phase Doppler Interferometer (PDI) is a useful tool to obtain at the same time the size and velocities of the burning kerosene droplets. In this paper, spatial distributions of these quantities for the steady spray flame, with comparisons between both operating points, are presented. A recirculation zone was evidenced from PDI results and the swirl motion of the flame was characterized. Even if the main goal of the study was to characterize the steady phase of combustion, particular attention was paid to the transient ignition phase, which is rarely addressed in the literature. In order to characterize spray and combustion instabilities experimentally, two different types of signals were analyzed and compared: high-speed laser planar images, and PDI data. Indeed, the time of arrival of the PDI data can be post-processed to reveal pulsations and clustering of spray droplets (Bachalo et al., 1990). In this paper, a fast Fourier transform was performed to detect cluster arrival frequency and periodicity of cluster formation. In such a spray flame, droplet clusters can be related to the periodic instabilities of the flame or pollutant formation

2 2. Experimental Setup 2.1 EPICTETE test Rig The EPICTETE swirl-stabilized combustion chamber is 1.4-m long and has a square cross-section of 100 mm x 100 mm (Brossard et al., 2010), as illustrated in Figure 1 Figure 1. The preheated air flow (m air =80 g/s and air temperature around 430 K) enters the combustion chamber through an industrial swirl-injector (Turbomeca Makila DLN). The kerosene (Jet-A1) spray, injected horizontally, along the X axis, by a DELAVAN nozzle, is formed at the tip (pilot flame configuration) of the swirl injector, which is representative to industrial conditions. The fuel-air equivalence ratio is ER=0.8. The burned gases exit the test rig through an exhaust. An adjustable needle can modify the chamber counter-pressure by changing the exit throat cross section. In this study, two conditions of chamber pressure were investigated: 100 kpa and 200 kpa. Optical diagnostics were performed through large quartz windows of 260 mm length and 100 mm height, placed on both sides of the combustion chamber. Measurements were performed under reacting conditions, within the stabilized period of combustion. Each fire test lasted 5 to 15 minutes. choked nozzle Turbomeca swirl injector window throttling plug exhaust combustion chamber Figure 1 : View of the EPICTETE test rig 2.2 Optical diagnostics High-Speed Laser Planar (HSLP) visualizations were performed in reacting flow conditions in order to define the area where the liquid phase is present. A Quantronix Darwin-Dual laser system was used to generate laser pulses shorter than 270 ns, at 527 nm and 4 khz repetition rate. A high speed camera (Photron APX-RS3000) recorded the Mie scattering signal from the spray droplets (see background image in Figure 2). A Phase Doppler Interferometer (PDI) from ARTIUM Inc., was used to characterize the kerosene droplet size and velocity distribution in the reacting flow, under steady operating conditions. A solid state laser system delivers green (λ g =532 nm) and blue beams (λ b =491.5 nm). The focal length of the emitter and receiver lenses was 500 mm. The off-axis angle of the receiver was set to 40 due to mechanical constraints around the test rig. The refractive index n of the kerosene droplets is temperature dependent, thus a value of n=1.4 was chosen between the preheated air flow conditions and the boiling state of the kerosene JP5 (Lefebvre, 1989), through the Eykman relationship (Pitcher et al., 1990). According to the PDI manufacturer, a temperature range from ambient to the boiling point results in an uncertainty of less than +/- 3% on the deduced droplet diameter. Based upon the optical setup, represented in Figure 3 the PDI could measure droplet diameters within the range 1.3 µm<d<192.8 µm, through the green channel. The number of validated droplet diameter measurements in each acquisition sample was at least 5,000, except for measurement - 2 -

3 locations where the droplet density was too low. The PDI provides two components of the droplet velocity (horizontal Vx and vertical velocity Vy, respectively through blue and green channel). The PDI instrument parameters were self-optimized to the velocity range based upon a 300 droplets sample analyzed before a measurement. The largest velocity range in our acquisitions was -176 m/s < Vx < 528 m/s and -166 m/s<vy<166 m/s for horizontal and vertical velocity, respectively. Z emitter Y O Vx X Vy 40 receiver Figure 2: locations of the PDI measurement points in the vertical plane O,X,Y. Background : instantaneous HSLP image of the spray. Figure 3: Top view of the PDI optical setup in the horizontal plane O,X,Z. Definition of PDI measured velocities. 2.3 Location of measurement points PDI measurements can be performed only where droplets are present. The two-phase flow area to be investigated was defined based upon time-averaged HSLP visualizations. An example of instantaneous HSLP image is presented in the background in Figure 2 with an inverted gray color scale. Measurement locations, illustrated by white dots in Figure 2, are located in three horizontal planes Y=0 and Y=+/- 15 mm. In these planes, transverse profiles were made along the Z axis, from X=5 mm to X=20 mm, with a 5 mm step. Diameter and velocity measurements near the nozzle exit (X=1.5mm and X=3mm) were also performed with the green channel only, thanks to the off-axis angle of the receiver. 3. Results 3.1 Droplet size measurements Two examples of probability density functions (pdf) are presented in Figure 4.a and b. They are based on the droplet size distribution by number and representative of the kerosene spray under reacting conditions, at P=200 kpa and ER=0.8. The corresponding measurement locations are along - 3 -

4 the injection axis, at X=12.5 mm, Y=Z=0 (Figure 4.a) and on the swirl cone envelope, at X=12.5mm, Y=-15mm, Z=0 (Figure 4.b). The pdf profile is fairly Gaussian in both cases, centered on 13µm (Figure 4.a) and on 19µm (Figure 4.b). The droplet density is the largest in the swirl cone envelope, about twenty times larger (2764 droplets/s, see Figure 4.b), than along the injection axis (128 droplets/s, see Figure 4.a). A few very large droplets (around 10 in the acquisition samples of size N, represented in Figure 4) were observed in the diameter range (50<D<200µm). Due to their low number, they are not visible in the droplet size distribution by number, but are responsible of a high D 32 compared to D 10. a) Injection axis : X=12.5mm, Y=Z=0, P=200kPa b) Swirl cone : X=12.5mm, Y=-15mm, Z=0, P=200kPa Figure 4 : droplet size distributions of the spray, reacting case, P=200 kpa. The spatial distribution of the Sauter Mean Diameter (D 32 ) is presented in Figure 5, for both operating conditions investigated (P=100 kpa left side and P=200 kpa on right side, same scale). In the background, corresponding time-averaged HSLP images are shown. The swirl cone angle, estimated through averaged HSLP images, is larger for the 200 kpa case, because the counterpressure results in a more compact flame. A general trend for Figure 5 is an increase of the droplet size radially towards the outer region of the spray. The smallest droplets are in the swirl core region, near the injection axis, and the largest droplets are on the outer borders of the spray. This spatial distribution is probably due to the presence of a recirculation zone which brings the unburned droplets back in the swirl core, increasing the residence time and droplets evaporation in this region (see Pascaud et al., 2005). Figure 5.b also shows that the droplet size decreases with the distance from the injector

5 Z axis (mm) X axis (mm) µm 35 Z axis (mm) X axis (mm) µm a) P=100kPa. b) P=200kPa. Figure 5 : Spatial distribution of D 32 in the horizontal plane Y=0. Dotted lines indicate the swirl cone angle estimated from time averaged HSLP images (background image). 3.2 Velocity measurements Axial velocity distributions (not shown here) are different in the recirculation zone than in the spray cone region. Indeed, in the central zone, a population of droplets with negative velocities (Vx - 50 m/s) is mixed with droplets of high axial velocity (Vx 150 m/s). Outside from the injection axis region (for Z ±5mm), only positive velocities are observed. Figure 6 shows the spatial distribution of axial velocities Vx in the horizontal plane Y=0, for the operating condition P=200 kpa. The spatial variations of Vx are stronger radially than axially, as observed for the mean droplet size in Figure 5. This trend is similar for the other operating condition P=100 kpa (not shown here). Negative axial velocities are observed in the region close to the injection axis, corresponding to 5 X 20 mm for P= 200 kpa (Figure 6) and 10 X 30 mm for P= 100 kpa (not shown here). For such a swirling flow, the transition from negative to positive axial velocities defines the stagnation point of the vortex breakdown, according to Lucca-Negro and O Doherty, The axial velocity reaches a maximum for measurement points in the swirl cone envelope, and a minimum in the injection axis. The locations of the maxima are in agreement with - 5 -

6 the dotted lines indicating the location of the swirl half cone angle, deduced from time-averaged HSLP images (see background images in Figure 5.b). We also noticed that the half-cone angle was slightly smaller in the region Z > 0. No particular size/velocity correlation was revealed from the PDI data analysis, in any location of the spray, neither on Vx nor on Vy Vertical Velocity Vy (m/s) Z axis(mm) X=1.5mm X=10mm X=20mm -100 Figure 6 : Spatial distribution of mean axial velocity Vx in the horizontal plane Y=0, P=200 kpa. Dotted lines indicate the 50 half cone angle of the swirl. Figure 7 : Radial profiles of the mean vertical velocity Vy (m/s) in the plane Y=0, P=200 kpa. Three radial profiles of the vertical velocity Vy are presented in Figure 7, for P=200 kpa. The swirl motion of the flow is clockwise, which is in agreement with the orientation of the nozzle vanes (see picture in Figure 7). The vertical velocity profiles from X=1.5 mm to X=20 mm show a global conservation of the rotating motion of the swirl. The same general trends are observed for both operating conditions (P=100 kpa and P=200 kpa), indeed the absolute maxima of the radial profiles decrease as X increases, and the position of these maxima is shifted outwards radially, thus indicating a small expansion of the swirl cone. The maxima of vertical velocity recorded in the radial profiles in the horizontal mid-section plane Y=0 correspond to the swirl tangential velocities. Therefore, the turnover time of the swirling flow could be estimated at τ swirl 1.35 ms based upon the radial locations of the maxima in the profile X=1.5mm close to the nozzle. By using the Vx and Vy radial profiles at the nozzle exit, the value of the swirl number S can be calculated. For the P=200 kpa operating condition, a value of S 0.65 was obtained, which is consistent with the results from Meier et al., 2007, obtained with the same injector but for methane combustion. The velocity profiles also allowed to characterize the axisymmetry of the spray. The spray was found to be fairly axisymmetric with respect to the injection axis

7 3.4 Study of the ignition phase The whole measurement campaign was performed through several fire tests, which naturally began by an ignition phase. As the PDI sampling rate is in the MHz range, it is possible to focus on data acquired during the ignition phase. In Figure 8, the temporal evolution of the diameter measurement of the first droplets, during the ignition phase, is presented for the measurement location X=20mm, Y=0, Z=0, P=100 kpa. The flame ignition was performed at ER 0.9, slightly above the nominal condition. Following two bursts of droplets (detected at t=11.5 s and t=13.5 s), the combustion began at t=14.2 s. The droplet diameters of the first detected droplets are close to the one in the combustion phase. During the ignition, as well as during the beginning of the combustion, a few droplets of large diameter (D 180µm) were observed. The values of the droplet size and velocities (not shown here) remain unchanged from the ignition phase to the steady phase of combustion. Figure 8: temporal evolution of droplets diameter in the ignition phase. X=20mm, Y=0, Z=0, P=100 kpa 3.5 Frequency analysis of instabilities Instantaneous HSLP images, recorded at 4 khz, revealed high spatial and temporal variations. A Fast Fourier Transform (FFT) was applied on the instantaneous mean gray intensity levels, defined by spatially averaging the intensity over the whole image. This analysis, performed over the steady phase, revealed a main instability peak at a frequency around 1kHz: khz for ER=0.8, P=100 kpa and khz for ER=0.8, P=200 kpa. For the P=100 kpa case, the first harmonic frequency is also visible (see Figure 9, top, dotted line). For the P=200 kpa case, an additional characteristic frequency of ~2000 Hz is also revealed by the FFT (see Figure 9, bottom, dotted line). However the validity of this 2000 Hz frequency peak deduced from the analysis of HSLP images is questionable, as it is only half of the 4 khz frame rate of the images. The FFT analysis was also conducted for the P=100 kpa case for various increasing equivalence ratios in the range 0.7<ER<1.2, and led to an increase of the main frequency from 820 Hz to 885 Hz. Moreover, for ER=0.8, the main frequency increased from 830 Hz to 1130 Hz when the - 7 -

8 chamber pressure was increased from 150 to 200 kpa. Therefore, these instabilities frequencies clearly depend upon the flame operating conditions. In an attempt to discriminate the frequency origin, the FFT analysis was also conducted locally, by spatially averaging the intensity over different regions of interest of 10x10 mm² in HSLP images, rather than in the whole image. Three regions were selected, centered on X=12.5 mm and Y=0 and Y=±15 mm. However, these local FFTs (not shown here) led to the same results. Figure 9: HSLP and PDI Fast Fourier Transform spectra for P=100 kpa (top) and 200 kpa (bottom), ER=0.8 The high sampling rate of PDI data (MHz range) was utilized to obtain the frequency content of the spray through FFT analysis. Different regions of the spray were investigated: in the swirl core region and in the swirl envelope. Because the PDI sampling is random, a re-sampling is required prior to FFT analysis. The re-sampling Refined Sample and Hold algorithm by Nobach et al., 1998, was applied through the ASSA software, developed at ONERA by Micheli et al., A 4 khz cut-off frequency limit was applied in the FFT computations. We applied the FFT on arrival times of axial velocity data Vx, corresponding to the green channel, which always exhibited a better data rate. Figure 9shows the FFT spectra obtained for three different PDI measurement points for each case investigated. These measurement points were located in the vertical mid-section plane (Z=0) at Y=0, +15 and -15mm; due to different swirl angles (see Figure 5), two different axial locations were chosen, X=20 mm, for the P=100 kpa case and at X=12.5 mm for the P=200 kpa case. No frequency peak was detected in the Hz frequency range, therefore this range is not presented in Figure 9. PDI revealed the same frequencies as HSLP images (f 1150 Hz for P=200 kpa and 850Hz for P=100 kpa), except for the measurement point located on the injection axis X=12.5mm, Z=0 and Y=0. From the PDI data analysis, the 2000 Hz frequency was only detected in the swirl envelope (X=12.5mm, Z=0 and Y=±15 mm, see Figure 9, and X=12.5mm, - 8 -

9 Y=0 and Z=±15 mm, not shown here) and not on the injection axis (X=12.5mm, Z=0, Y=0). Therefore, this 2 khz frequency could be specific to the swirl region. PDI FFT analysis also revealed frequencies in the range Hz (blue, red and green solid line peaks, Figure 9, bottom) for the P=200 kpa case, that are not detected in the FFT analysis of the HSLP images. In the operating conditions investigated in this paper, the swirling flow generated by the nozzle is characterized by a swirl number S of 0.65 (see section 3.2). According to Lucca-Negro and O Doherty, 2001, such a high swirl number can induce an axial vortex breakdown, associated with a Precessing Vortex Core (PVC). Pascaud et al., 2005, estimated the PVC characteristic frequency through the relationship f=1/τ swirl. In the present experiment, the turnover time was found to be τ swirl 1.35 ms for the P=200 kpa case investigated in this paper (see section 3.2). Therefore, it provides an estimation of the characteristic frequency of f 743 Hz. On one side, this value is close to frequencies in the range Hz detected in the FFT analysis of the PDI signals (blue, red and green solid line peaks, Figure 9, bottom), but not in the FFT analysis of the HSLP images. This could be explained by a much stronger sensitivity of the PDI to PVC motion due to the very high spatial resolution of this technique (very small measurement volume). On the other side, Poinsot and Selle, 2005 characterized the PVC numerically for non-reacting and reacting swirling flow conditions, and indicated inhibition of the PVC by the reacting flow. Thus the aerodynamic origin of these frequencies has not been clearly identified yet. The instabilities are not attributed to a coupling with the acoustic modes of the chamber. Indeed, considering the operating conditions at P=100 kpa and the combustion chamber dimensions, the first resonant acoustic modes are 408 Hz (first longitudinal mode), 4487 Hz (first transverse modes) and even higher for tangential modes. The value of 408 Hz does not match the frequencies evidenced by PDI or HSLP images. In order to fully identify the phenomena involved with the instabilities evidenced in this paper, additional experimental data are required, such as dynamic pressure sensors placed at different locations in the chamber walls, and high-speed OH* chemiluminescence images. 4. Conclusion The droplet size and velocity distributions in a swirl-stabilized kerosene/air combustor representative of aeronautical combustors were investigated by Phase Doppler Interferometry. Two different pressure conditions, P=100 kpa and 200 kpa, were studied. Spatial profiles of droplet size and 2-components velocity measurements were analyzed in three horizontal planes and close to the nozzle exit. Droplets were found to be slightly larger in the swirl cone envelope than along the injection axis, due to the presence of a central recirculation zone, evidenced by negative axial velocities measured along the injection axis. The axial and vertical velocities reach a relative maximum for measurement points in the swirl cone envelope and a minimum in the injection axis region. The locations of the maxima are in agreement with the location of the swirling spray cone visible in the averaged planar laser sheet image of the spray. Comparisons of the results obtained for the two operating conditions investigated showed similar trends for the spatial profiles of velocities and diameters. However, slightly smaller droplet sizes, as well as and a smaller half-cone angle of the swirl, were observed for the 100 kpa case. The high frequency sampling of PDI data allowed to study the ignition phase and the frequency content of instabilities. This frequency content was found to be consistent with the one deduced from high-speed laser images. Both techniques showed frequencies in the khz range, the exact values of which depended upon the experimental conditions. In order to fully identify the phenomena involved with the instabilities evidenced in this paper, additional experimental data would be required from a future work, such as dynamic pressure sensors placed at different locations in the chamber walls, and high-speed OH* chemiluminescence images

10 5. References Bachalo W., Rudoff, C.R. and Sankar, S.V. (1990), Time-resolved measurements of spray drop size and velocity, In: Hirleman, Bachalo, Felton Eds., Liquid particle size measurement techniques, 2 nd volume, American Society for Testing and Materials, pp Brossard C., Vincent-Randonnier A., Barat M. and Gicquel P. (2010), Caractérisation par PIV haute cadence d une combustion stabilisée par injection tourbillonnaire, CFTL 2010, Vandoeuvre-lès- Nancy. Lefebvre, A.H. (1989), Atomization and Sprays, Hemisphere publishing Corporation. Lucca-Negro and O Doherty (2001), Vortex breakdown: a review, Prog. Energy Comb. Sci. Vol.27, pp Meier W., Weigland P., Duan X.R., Giezendanner-Thoben R. (2007), Detailed characterization of the dynamics of thermoacoustic pulsations in a lean premixed swirl flame, Combustion and Flame, vol.150, pp Micheli F., Lavieille M., Millan P. (2006), ASSA, un outil de référence pour le traitement du signal en vélocimétrie laser, 10ème Congrès Francophone de Techniques Laser (CFTL), Toulouse, France. Nobach H., Müller E.and Tropea C. (1998), Correlation estimator for two-channel, non-coincidence laser-doppler-anemometer, 9th International Symposium on Applications of Laser Techniques to Fluid Mechanics, Lisbon, Portugal. Pascaud S., Boileau M., Martinez L., Cuenot B. and Poinsot T. (2005), Large Eddy Simulation of turbulent spray combustion in aeronautical gas turbines, ECCOMAS Thematic Conference on computational combustion, Lisbon, Portugal. Pitcher, G., Wigley, G., and Saffman, M. (1990), Sensitivity of dropsize measurements by phase Doppler anemometry to refractive index changes in combusting fuel sprays, Fifth International Symposium on Applications of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, pp Poinsot T.and Selle L. (2005), LES and acoustic of analysis of combustion instabilities in gas turbines, In PLENARY LECTURE - ECCOMAS Computational Combustion Symposium, Lisbonne, Portugal. Acknowledgment This work was performed with financial support from the French DGA (Délégation Générale de l Armement) and ONERA