An Experimental study of Coupling between Combustor Pressure, Fuel/Air Mixing, and Flame Behavior

Transcription

1 An Experimental study of Couplin between Combustor Pressure, Fuel/Air Mixin, and Flame Behavior D. M. Kan, F. E. C. Culick Jet Propulsion Center/Department of Mechanical Enineerin California Institute of Technoloy Pasadena, California and A. Ratner Department of Mechanical & Industrial Enineerin University of Iowa Iowa City, Iowa Abstract Fuel-air mixin behavior under the influence of imposed acoustic oscillations has been studied by investiatin the response of the fuel mixture fraction field. The distribution of local fuel mixture fraction inside the mixin zone, which is expected to evolve into the local equivalence ratio in the flame zone, is closely coupled to unstable and oscillatory flame behavior. As part of a series of works to study the combustion dynamics, an experiment was performed with an aerodynamically-stabilized non-premixed burner. Acoustic oscillations were imposed at, 7, 3, 37, and 55Hz. Phase-resolved acetone PLIF was used to imae the flow field of both isothermal and reactin flow cases and this data alon with the derived quantities of temporal and spatial unmixedness were employed for analysis. The behavior of the unmixedness factor is compared with the previous measurements of oscillations in the flame zone. This comparison suests that local oscillations in fuel/air mixin are closely related to the oscillatory flame behavior. For each drivin frequency, the mixture fractions oscillate at that frequency but with sliht phase differences from the combustor pressure/flame intensity, indicatin that the fuel mixture fraction oscillation are likely the major reason for oscillatory behaviors of this cateory of flames and combustor eometry. Introduction Combustion instability is a phenomenon where acoustic waves form and et amplified inside a combustor and induce oscillatory flame behavior. The unstable burnin often results in a reduction of combustor lifetime and a decrease in performance. Combustion instabilities also cause inhomoeneous burnin, which can result in elevated pollutant levels and produce unacceptably hih levels of vibrations and heat transfer rate, leadin to structural damae. These conditions were typically avoided by operatin combustors in a stoichiometric or fuel-rich-pilot confiuration. Due to increasinly strinent uidelines on pollutant (NOx) production and the desire for hiher fuel efficiency, there has been recent wide-spread movement to employin lean premixed combustion schemes for industrial dump combustors. But, as is well know, these schemes have an inherent tendency towards producin combustion instabilities. Several research efforts have been devoted to the observation and measurement of unstable combustor pressure fluctuations, and also to the control of this phenomenon [1-4]. Mostly, the efforts to actively control the combustion process have been case-specific. They did not address the need for a eneral, extensive study of the mechanisms leadin to combustion instabilities nor were able to elucidate the knowlede basis of eneral combustion dynamics. In an effort to establish a solid knowlede base for combustors desin and a eneralized methodoloy for the active control of combustion instabilities, a project was undertaken to accurately measure the combustion dynamics of a complex, laboratory-scale system. By measurin various aspects of flame behavior, fuel/air mixin, and other factors, physical and numerical models of the combustion processes can be constructed. When enouh data has been collected over a wide rane of conditions, models of combustion response can then be developed. These models will provide new insihts into the dominant physical processes, a means to better estimate the behavior of newly developed combustors while still at the desin phase, and techniques for the control of combustion instabilities for those systems that are already in operation. As part of this research effort, oscillatory behavior of a non-premixed burner was studied by Pun et al. [5, 6]. These studies examined the flame dynamics under forced acoustic oscillations, and showed the phase-dependent responses of the combustion process for certain acoustic excitations frequencies (-55Hz). The current work is a study of the behavior of fuel/air mixin under these same conditions. Lieuwen et al. [8] showed, theoretically, that the manitude of the reaction rate and heat release oscillations produced by equivalence ratio perturbations increases sinificantly (by a factor of 5-100) as the equivalence ratio (unmixedness) decreases, especially under leaner burnin condition. The effect of temperature sinificantly decreased while the effect of flow rate perturbations was constant and did not show any amplification. This implies that the equivalence ratio perturbations play a key role in drivin combustion instabilities in fuel lean environments. The current work presents measurements, obtained by means of phase-resolved acetone PLIF (employin standard techniques) [9-1], of how the fuel/air mixin fluctuates in

2 response to imposed acoustic oscillations. Also, by comparison with the previous measurements of Pun et al. [5, 6], the oscillatory behavior of the flame can be related to the mixture fraction oscillations. By comparin the phasin of the drivin acoustic field to both the fuel mixture fraction oscillation and the flame itself, the strenth and direction of physical couplin can be determined. Experimental Confiuration The burner, as shown in Fiures 1 and, is a traditional jet-mixed type burner with flame anchorin occurrin approximately in the middle of the quartz tube, with the exact heiht dependin on the specific flow conditions such as fuel/air ratio. The fuel jet is 50% methane and 50% nitroen and air is entrained and drawn into the jet as the flow moves throuh the eductor. The quartz tube is 5.7cm wide on each side and 11.43cm tall and is made of fused silica/quartz to enable observation and measurement of the flame. Acetone was seeded into the fuel stream as a marker and was fully mixed with fuel before injection into the combustor. Acetone PLIF was then used to measure the distribution of fuel in the mixin reion upstream of the flame zone (as shown in Fiure 1). function enerator provide the input sinal and power to the speakers. Fiure. The combustion chamber: (a) loud speakers, (b) pressure transducer, (c) fused-silica burner tube, (d) eductor block (see Fi. 1), (e) fuel spud, arrows at the bottom indicate the air inlet. Fiure 3 is the layout of the acetone PLIF imain system. An intensified CCD camera is used for the imae acquisition, while a National Instrument data acquisition board (NI PCI 6014) alon with pressure transducer (PCB 106B50) is used to measure and record the pressure and other sinals. PLIF imain of acetone is performed at the bottom portion of the quartz tube where no flame is present, as indicated in Fiure 1. Fiure 1. Fuel mixture fraction is measured with acetone PLIF in reion (). A schematic view of the acoustic chamber is shown in Fiure. The acoustic chamber is made of an aluminum central section and a stainless steel top that houses the loudspeakers and allows direct exhaust of the product ases at the top. The exhaust port is unconstrained and is open to the atmosphere. At the bottom of the chamber, a circular section with radial vents allows air to enter but creates a closed-end acoustic condition. This section has two sets of inlet louvers cut on opposin sides to allow this radial airflow into the chamber, while maintainin acoustic closure. This creates a closed-end bottom and open-end top acoustic conditions. The loudspeakers housed in the upper portion of the chamber are used to enerate the acoustic field. To protect the speakers from heat failure, they are attached to an air-jet film-coolin system. The loudspeakers are 1 inch (30 cm in diameter) and can handle 400 W (each) of continuous power. A 1000 W power amplifier alon with a Fiure 3. Schematics of the acetone PLIF. The ND:YAG laser has internal laser frequency doublin and outputs a nd harmonic hih power beam at 53 nm. This beam is used to pump a dye laser which operates at 560

3 nm. The output of the dye laser is, in turn, frequency doubled to 80 nm for excitation of acetone. The laser power enterin the test section was 8.4 mj/pulse, which is an intensity of approximately mj/pulse/cm. The PLIF sinal is captured on an intensified CCD camera with a maximum resolution of 51 by 51 pixels. An area of 5.5 by 4.1 cm is imaed onto 300 by 5 camera pixels. The PLIF sinal passes throuh a UV hih-pass filter which blocks all liht lower than 300 nm in wavelenth. This blocks laser beam scatter and passes the fluorescence sinal which occurs mostly between 350 and 550 nm. Imaes are taken at random pressure phases, with the camera atin sinal bein recorded by the data acquisition system alon with the pressure sinal so as to enable appropriate post-processin. Post-processin involves sortin the imaes by the pressure phase, creatin phase-averaed imaes, normalizin based on laser intensity. Results The lobal unmixedness is defined as σ U =, (1) (1 < x > ) ( < x > ) where σ is the standard deviation and <x> averae of fuel concentration over the entire -D imae, instead of one point with many measurements. The unmixedness is a normalization [13] of the variance σ by the maximum possible value for the iven <x>, evaluated by the variance of Housdorf relation, σ max =< x > (1 < x > ). When the fuel is completely mixed and homoeneously distributed, U is zero; when no mixin occurs, U is unity. A twodimensional imae collapses to a sinle value by this definition, which ives a quantitative measure of the manitude of the variation of fuel/air mixin, the deree of fluctuation in fuel concentration over the entire reion. The local temporal unmixedness is defined as in the work by Fric [14], σ t Ut =, () (1 < x > t ) ( < x > t ) where σ t is the standard deviation of the fuel concentrations drawn from repetitive measurements at the specific location over time, and <x> t is the time averae of fuel concentration at that location. In respect to temporal unmixedness (U t ), hiher values of U t mean reater fluctuations in the fuel concentration. Thus, where U is larer can be interpreted as a reion where more mixin occurs than in locations with lower values of unmixedness values. It is therefore expected that the hih unmixedness reion becomes wider down stream with the flow. It is shown that hih values of the temporal unmixedness occur in the shear mixin layer of the flow (30-75% from the center, see Fi 4). There is not much variation between cases at different excitation frequencies, but they show clear position-wise dependency. In fiure 4, lihter shadin indicates a reion with a hiher value of unmixedness, thus markin a reion of hih fuel concentration fluctuation. Further downstream from the eductor block (upward in Fiure 4), the shear mixin zone widens as expected. The low temporal unmixedness in the core reion, of 0 30 % distance from the center, is due to the hih, and relatively uniform, fuel concentration. This shows the core reion is still fuel dominant and needs further mixin Here, the dark reion outside the shear mixin layer is larely due to homoeneously low fuel concentration. Fiure 4. Two-dimensional maps (adjusted) of temporal unmixedness at 37Hz, (top) reactin and (bottom) nonreactin cases. In the reactin case, due to the turbulent flow field enerated by the heat release from the combustion processes, the unmixedness values in eneral are much hiher than those in the cold flow as well as in the buoyancy driven flow structure in the reactin case. The reactin cases have a flow with sinificant post-reaction buoyancy, creatin an enhanced vertical velocity component. This means that for the same vertical location, the reactin flow cases will resemble lower vertical locations (earlier times) in the nonreactin cases. This is evidenced by the fact that the nonreactin cases are mostly mixed by the time the flow reaches the imain area while the reactin flows are still underoin mixin (Fiure 4). Fiure 5 is a 3D plot of the power density spectrum [7]: radial location vs. the Fourier transform of the time series 3

4 data at each location vs. intensity of each frequency at each radial location. This particular plot is for a 3 Hz driven frequency and reactin flow, and shows that there is a very stron peak in the mixture fraction oscillations in response to an imposed acoustic field. This spike is not only narrow in frequency, but also matches the drivin frequency as the frequency was varied from to 55 Hz. The spike was between and 3 orders of manitude stroner than the natural, low frequency oscillations that are present. This clearly demonstrates that the acoustic oscillations cause oscillations in the fuel mixture fraction in the pre-flame zone at each drivin frequency and, that there is a stron complex couplin between flame oscillations, and the resultin acoustic field. Fiure 7 shows the oscillatory behavior of mixin at frequencies of 7 and 37 Hz, with and without the flame present. The sine waves in bold lines correspond to the imposed acoustic waves. Pressure waves, assumed to be in pure sinle-frequency sine wave form, are used as the reference for determinin the phase of the data collected. The measured oscillatory mixin behavior is shown with the approximate sine waves correspondin to the first mode of unmixedness oscillations (blue lines). Examinin the trendwise behavior, the deree of mixin (unmixedness) varies (in an oscillatory fashion) at the drivin frequency (Fiure 5) with some phase differences, even thouh there s a sliht difference in phase and manitude of the mixin oscillations. Fiure 5. Power density spectrum for 3Hz, reactin flow [7]. Since no stron frequency preference is seen in the outer portion of the flow, it is most likely that the couplin is not a stron function of the vortex sheddin from the fuel tube exit. The phase dependence of the mixin behavior is shown in Fiures 6 and 7. The role of combustion processes in causin hiher uncertainties in fuel/air mixin is observed from the comparison with the non-reactin case (blue, in Fiure 6), where the only difference is the presence of flame. The lobal unmixedness is presented versus the excitation frequency and the phase durin a cycle of excitation at each frequency in Fiure 6. Greater values in lobal unmixedness factor, the deree of inhomoeneous mixin in the reion, is observed in the presence of flame, and the mixin at hiher frequencies (3-55Hz) is much more affected by the presence of the flame than at lower frequencies (, 7Hz). The combustion process alone causes hue differences in mixin with increase in the unmixedness value up to a factor of 4. The increase in unmixedness with frequency for the reactin flow case seems to be caused by the interaction between the combustion process and the acoustic excitation, and it seems more plausible when compared to the non-reactin case where the tendency is actually in the opposite direction. While combustion process has a reat effect on the mixin behavior, mixin is also affected by the phase of excitation. 4 Fiure 6. 3-D representation of lobal unmixedness vs. excitation frequency and phase. Measurements show quite clearly that the mixture fraction oscillates at the same frequency as the acoustic drivin frequency. This suests that either both the mixture fraction oscillations and the flame oscillations are coupled directly to the acoustic forcin or that the system has three-way couplin where (in addition to the acoustic couplin) the presence of oscillation in mixture fraction induces oscillations in the flame and/or oscillations in the flame induce or enhance mixture fraction oscillations. Lieuwen et al. [8] arues that the oscillations in local mixin induce or strenthen oscillatory behavior in flame burnin, as is seen here. Conversely, the non-reactin flow cases show much weaker mixture fraction oscillations than the reactin cases, implyin that flame oscillations enhance mixture fraction oscillations. This would imply a bidirectional couplin. It must be noted that, in this work, there are clear limitation in bein able to compare to or support Lieuwen et al. s conclusions due to that, in this work, the oscillatory fuel/air mixin is not the input variable bein modulated as it is in Lieuwen et al. s work [8] but is another output, just as is flame behavior, resultin from the imposed acoustic excitation. Fiure 8 shows a comparison of the phase shifts between the flame, the mixture fraction in reactin cases and the mixture fraction in non-reactin cases. There is a

5 clear phase differences between reactin and non-reactin flow cases but the phase chane with drivin frequency follows the same trend for all three variables. When the differences in phase behavior between the flame and the mixture fraction are examined, especially at frequencies 7- consistent. Such uniformity with chanin frequency implies a chemical time-scale that is independent of drivin frequency, but nothin can be concluded without further investiation. Fiure 7. Oscillatory behaviors of lobal mixin presented with reference acoustic wave (dark sine wave). Thin blue lines indicate the approximation of mixin behavior to phase-shifted sine waves. Fiure 8. Comparison of phase shifts of mixin and flame [5, 6] behaviors. Phase shift from the drivin acoustic wave (left); from the oscillations in flame (riht). 55Hz (for reactin flow cases only), the phase difference is consistently derees (lain) or, derees (leadin). It is not clear why the phase differences are so Conclusion Mixin is a sinificant factor in causin unstable combustion in systems that operate in a premixed or 5

6 partially premixed combustion confiuration and where the mixin occurs in the combustion chamber. Onset of unstable combustion for these systems is due to the presence of mixture fraction oscillations which are in turn the result of chamber acoustic oscillations. These systems exhibit stron couplin between mixin and the imposed acoustic field, visible in the twodimensional temporal unmixedness distribution. The distribution map provides information quantifyin how fuel/air mixin is structured in the mixin reion, and where the fluctuations in temporal unmixedness are stronest. It is clear from the results that the acoustic forcin causes a stron periodicity in the mixin layer at the driven frequency. In this work, the effect of the oscillations on both mixin (unmixedness) and mass flow rate could not be decoupled due to the limitations inherent in this experiment. Still, the acoustic-waves-induced oscillations in the overall mixin were clearly evident at all drivin frequencies. The presence of a flame intensifies the oscillations present in the mixin zone, in terms of temporal unmixedness, and results in more structure and sharper radients for mixture fraction than that found in the non-reactin cases. The heat release and induced buoyancy produced by the combustion process also play important roles in enhancin acoustic field-mixture fraction couplin. Acknowledements This work was supported in part by the California Institute of Technoloy and partly by the Air Force Office of Scientific Research (AFOSR) under Grant No. F (Dr. Mitat Birkan, Proram Manaer). We thank Carlos Pinedo for assistance in confiurin the experiment. d Influence of Oscillations in Fuel Mixture Fraction on F lame Behavior, Proceedins of the 3rd Joint Meetin of the Combustion Inst., Lieuwen, T., Neumeier, Y., and Zinn, B.T., The Role of Unmixedness and Chemical Kinetics in Drivin Combu stion, Combust. Sci. and Tech., 135:193-11, Thurber, M.C., and Hanson, R.K., Simultaneous imain of temperature and mole fraction usin acetone planar laser-induced fluorescence, Exp. of Fluids, 30:93-101, Demayo, T. N., Leon, M.Y., Samuelsen, G. S., and Hol deman, J. D., Assessin Jet-Induced Spatial Mixin in a Rich, Reactin Crossflow, J. of Prop. and Power, 19(1), 14-1, Yip, B., and Miller, M. F., "A Combined OH/Acetone P LIF Imain Technique for Visualizin Combustin Flo ws." Exp. in Fluids, 17(5): , Lozano, A., Yip, B., and Hanson, R.K., Acetone : a Tracer for Concentration Measurements in Gaseous Flows by Planar Laser-Induced Fluorescence, Exp. Fluids, 13: , Dimotakis, P.E., and Miller, P.L., Some Consequences of the Boundedness of Scalar Fluctuations, Phys. Fluids A (11): , Fric, T.F., Effects of Fuel - Air Unmixedness on NOx Emissions, J of Prop. Power, 9(5): , 1993 References 1. Dowlin, A.P., Vortices, Sound and Flames a damai n combination, The Aeronautical Journal, pp , Venkataraman, K.K., Preston, L.H., Simons, D.W., Lee, B.J., Lee J.G., and Santavicca, D.A., Mechanism of co mbustion instability in a lean premixed dump combusto r, J of Prop Power, 15(6): , Mohanraj, R., Neumeier, Y., and Zinn, B.T., Combustor Model for Simulation of Combustion Instabilities and Their Active Control, J of Prop. Power, 16(3): , Paschereit, C.O., Gutmark, E., and Weisenstein, W., Excitation of Thermoacoustic Instabilities by Interaction of Acoustic and Unstable Swirlin Flow, AIAA Journal, 38(6): , Pun, W., Palm S.L., and Culick, F.E.C., PLIF Measure ments of Combustion Dynamics in a Burner under Force d Oscillatory Conditions, 36th AIAA/ASME/SAE/ASE E Joint Propulsion Conference and Exhibit, AIAA , Pun, W., Palm, S.L., and Culick, F.E.C., Combustion d ynamics of an acoustically forced flame, Combust Sci a nd Tech., 175: , Fernandez, V., Ratner, and A., Culick, F.E.C., Measure 6