Measurements of Hydrogen Syngas Flame Speeds. at Elevated Pressures

Transcription

1 Paper # A16 Topic: Laminar Flames 5 th US Combstion Meeting Organized by the Western States Section of the Combstion Institte and Hosted by the University of California at San Diego March 25-28, 27. Measrements of Hydrogen Syngas Flame Speeds at Elevated Pressres M.P. Brke, 1 X. Qin, 1 Y. J, 1 and F.L. Dryer 1 1 Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 8544, USA Laminar flame speeds for syngas mixtres of varios compositions p to 2 atm are measred with otwardly propagating spherical flames at constant pressre sing more restrictive measres for data redction that consider deviation from the ideal flow field cased by a non-spherical confinement. In the present paper, as in earlier high pressre stdy of flame speeds, a recently developed constant-pressre approach that tilizes a cylindrical test chamber is applied. For the first time, the effect of flow field on flame speed measrements is addressed in detail for these non-spherical constant-pressre chambers. The reslts from experiments and simplified analysis indicate that deviation from the assmed flow field cases significant errors in instantaneos flame speed measrement, which are amplified in the extrapolation to zero stretch rate. The relative deviation of the apparent flame speed from the tre flame speed is fond to scale with (σ-1) times the relative flow field deviation, where σ is the nbrned to brned gas density ratio. This reslt demonstrates the significance of the flow field on the measred flame speed, since the density ratio typically assmes vales of ~5-9 for hydrogen and hydrocarbon fels in air. A simple model is developed to stdy the effect of flow distrbance in cylindrical confinements. In cylindrical chambers, where the flow is typically most constrained in the plane of measrement (radial direction), failre to consider this effect reslts in lower vales for the measred flame speed. Laminar flame speed measrements are reported for H 2 /CO/CO 2 mixtres varying in eqivalence ratio from.6 to 4., pressre from 1 to 2 atm, and CO 2 diltion from to 25%. The corrected data range (.6 cm < r <.3R w where R w is the chamber wall radis) is seen to raise the extrapolated brning velocity by as mch as 6% from the brning velocity fond from extrapolation over a wider range (.6 cm < r <.54R w ), which is more strongly inflenced by flow field deviations. The experimental measrements are compared with experiments from McLean et al., Sn et al., and Hassan et al. (where applicable) and planar calclations sing the kinetic mechanisms of Li et al., Davis et al., and Sn et al. While the experimental data and predictions for brning velocity agree reasonably well at lean conditions, large discrepancies occr at rich conditions. The sbstantial variation in the available stretch-corrected flame speed data indicate that significantly more data and more rigoros calclations of ncertainty are necessary before qantitative conclsions can be made and the performance of kinetic mechanisms can be properly assessed. 1. Introdction The planar brning velocity is often termed a fndamental combstion parameter, providing a measre of the copled effects of exothermicity, diffsivity, and reactivity of a combstible mixtre. It is the only flame speed niqe to a gas of a fixed composition, initial temperatre 1

2 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames and pressre, withot frther specification of hydrodynamic conditions, sch as stretch rate, Reynolds nmber, extent of wrinkling, etc. The laminar flame speed serves as a sefl inpt parameter for laminar flamelet calclations for trblent combstion [1]. Not only does the brning velocity give information abot the way a flame behaves nder specific conditions, bt it can be sed to validate chemical kinetic and transport models. Conseqently, the planar brning velocity is a freqently soght parameter in the combstion commnity, despite the fact that actal nstretched flames are rarely (if at all) encontered in applications and they are relatively difficlt to measre. Reliable experimental data and validated kinetic and transport models for H 2 /CO combstion over a wide range of pressres, temperatres, compositions, and diltions are necessary for the development of technology for coal-derived syngas combstion in stationary gas trbine engines as well as for solidifying the fndamental bilding block for modeling high-pressre, hightemperatre hydrocarbon combstion. Brning synthetic gas is challenging in practice, becase its diverse composition, consisting of CO, H 2, CO 2, H 2 O, CH 4 and varios trace species, can prodce significant variations in combstion properties for different relative concentrations of its components. Therefore, kinetic mechanisms will be sefl for predicting combstion characteristics of synthetic gas mixtres of all possible compositions and mst be validated over a wide range of eqivalence ratios, pressres, temperatres, and species concentrations. Since most practical applications inclding gas trbine engines perform at high pressres, reliable experimental data and validated models for high pressre conditions are especially important. While a considerable amont of work on low pressre flame speeds for syngas mixtres has been performed [2-5], only one stdy has reported high pressre flame speeds [6]. Therefore, frther investigation of laminar flame speeds for higher pressres typical of practical applications contines to be of interest. The experimental method sed in this stdy for flame speed measrement is the constantpressre otwardly propagating spherical flame method. The spherical flame and conterflow flame methods are the two most common recently employed techniqes, becase their clearly defined stretch rates allow for extrapolation to an nstretched, adiabatic planar flame speed. New techniqes have been developed to extend the spherical flame method to achieve measrements at pressres significantly higher than those sed previosly [7,8]. While flame speed measrements have previosly been limited to pressres less than 1 atm, this method has allowed for stretch-corrected flame speed determination p to 6 atm [9]. The constant-pressre spherical flame method involves captring schlieren pictres of a spherical flame and calclating the instantaneos flame speed and stretch from radis-time data [1,11]. Most stdies employ the relations given by Strenlow and Savage [12] for an nconfined otwardly propagating spherical flame, in which the brned gas is assmed to come to rest after crossing the flame and the flame is taken to be infinitesimally thin, ρ dr b f s = (1) ρ dt drf κ = 2 (2) r dt f 2

3 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames where s is the stretched flame speed, ρ and ρ b the nbrned and brned gas densities, r f the flame radis, dr f /dt the flame propagation speed, and κ the stretch rate. The density ratio is freqently compted by adiabatic calclation sing the algorithm of Gordon and McBride [13], bt in this stdy it is calclated in a planar flame calclation [14] sing the mechanism of Li et al. [15]. Planar flame calclations sing the mechanism of Davis et al. [16] yielded density ratios within.5% of those fond sing Li et al. for all cases stdied. There are varios methods fond in the literatre for relating the stretch rate and the stretched flame speed sch that the fndamental mixtre parameters, the nstretched brning velocity, s o, and Markstein length, L, can be extracted as described in Ref. [17-21]. The present stdy employs the relation first postlated by Markstein [17] o s s L κ (3) = derived from asymptotic theory for weakly stretched flames [22], bt has been fond to hold for more general conditions as discssed in Ref. [23]. Finally, the nstretched flame speed and the Markstein length can be extracted throgh a linear regression of (3). It shold be mentioned that in stagnation and conterflow flame stdies a nonlinear extrapolation method has been developed to relate the flame speed at the experimentally determined reference srface, the location of minimm velocity, to the flame speed at the theoretical reference srface, the location of 1% temperatre rise [24]. As sch, the nonlinear extrapolation approach has no bearing on spherical flame stdies. For the aforementioned spherical-flame theory to be satisfied exactly, it reqires an nwrinkled, spherical, infinitesimally thin flame, which obeys the linear stretch relation, propagating in an nconfined environment with a constant density ratio across the front and no ignition, radiation or gravity effects. Of corse, these assmptions are never satisfied exactly and, conseqently, a large nmber of stdies have been devoted to qantification and correction of the errors incrred by departres from the theoretical assmptions, inclding finite flame thickness [25], ignition distrbances [2,26], pressre rise [27-34], varying density ratio [18], radiation [2,18] and flame front wrinkling [35-37]. In the present paper, as in earlier high pressre stdy of flame speeds, a recently developed constant-pressre approach that tilizes a cylindrical test chamber is applied [7,8]. We describe departres of experimental reslts from expected behavior that cannot be explained by any of the above corrective analyses. We discovered the effect of a non-niform flow field reslting from a non-spherical chamber, which has not been considered previosly. While previos stdies have mentioned the possibility of wall interference in these constant-pressre cylindrical bombs at large radii [9], this effect has not been qalitatively explained or qantitatively investigated. The motivation for investigating this effect in the present paper is that the cylindrical chamber geometry of the constant-pressre method sed here and in earlier papers makes the presence of this particlar effect more apparent, as shown below. In the remainder of this paper, we will describe the experimental method, investigate the effects of flow field non-niformity on the measrements, and report measrements on flame speed for syngas mixtres at high pressre with consideration of this effect. 3

4 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames 2. Experimental Methods The experiments are condcted in a dal-chambered, pressre-release type high-pressre combstion apparats shown in Figre 1. The complete details of the experimental apparats and procedre can be fond in Ref. [8]. The chamber consists of two concentric cylindrical vessels of inner diameters 1 and 28 cm. The length of the inner chamber is cm. Twelve holes of 2.2 cm in diameter are located in the lateral wall of the inner vessel to allow for pressre release. These holes are sealed with O-rings nder the compression of the iron plates attracted by a series of permanent magnets imbedded in the wall. These iron plates can provide a seal between inner and oter chambers with a pressre difference above.3 atm. When the pressre difference vanishes or reverses, the iron plates open allowing the gas to flow from inner to oter chamber to maintain a constant pressre. Frthermore, since the volme of the inner vessel is 1 times smaller than that of the oter, the total pressre increase after combstion is small enogh to ensre a constant pressre experiment and operational safety at high pressres. Qartz Window Focs Lens Pressre Sensor & Gages Oter Chamber Gas Releasing Holes High-Speed Video Camera Mercry Lamp Pinhole Collimating Lens Knife Edge Permanent Iron Plate magnet Decollimating Lens Tngsten Wire Inner Chamber Figre 1: Schematic of pressre-release type cylindrical bomb and schlieren configration Schlieren photography is tilized for imaging the flame propagation. Light from a 1 W mercry lamp is focsed on a 1 µm pinhole and collimated by a spherical lens. The collimated light passes throgh windows to the inner chamber and is focsed on a horizontally installed knife edge. A high-speed digital video camera with 4 µs shtter speed and frame rate of 24, fps is sed to record images of the propagating flame. From these images, the flame front is located and recorded sing an atomated detection program for ease of processing and redction of hman bias. The raw flame radis data are smoothed and flame front velocity calclated sing local 2nd order polynomials fit by least-sqares methods over a range of 3 mm srronding each point. The data processed in this manner are consistent with the raw data as well as data processed throgh a more conventional local-averaging (low-pass) filter. The polynomial fitting is performed for each point over short spans of radis-time data, bt not for the entire data set, which has been observed to prodce false trends. The local polynomial processing is sed to dampen any end effects of the regression analysis for stretch extrapolation and, more importantly, to aid in interpreting data observed in the experiment. 4

5 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames The plot of flame speed vs. stretch for a hydrogen-air mixtre of eqivalence ratio 3. at 1 atm shown in Figre 2a is indicative of the trend nder the conditions sed in this paper. There is a linear relation for large stretch (small radii), a change in crvatre for moderate stretch (moderate radii), and a sharp slope for small stretch (large radii). Flame speed (cm/s) cm Stretch (1/s) Extrapolated planar brning velocity ( cm /s) Starting radis (cm) (a) (b) Figre 2: Exemplar experiment for a hydrogen-air mixtre of eqivalence ratio 3. at 1 atm The reslting overall trend is not consistent with the theory sed to extract planar vales. However, none of the corrective analyses presented in the literatre address this deviation. The same trend is observed when the inner chamber is open or closed to the oter chamber. In fact, the data sed in this figre were taken when the iron plates were removed from the large holes connecting the inner and oter chambers. Therefore, the flame speed deviation in Figre 2 is clearly not cased by chamber pressre rise. Additionally, previos stdies indicate the inflence of radiation is small for these conditions [23]. Inspection of the schlieren movies reveals that the flame front image remains smooth and circlar throghot its propagation. In fact, the flame speed has been observed to obey a linear relation with stretch rate in experiments [38] and in direct nmerical simlations [34] nder similar conditions. It trns ot that the above departre of observations and theory is cased by a deviation from the assmed flow field cased by the cylindrical chamber geometry. From Figre 2b, it is also clear that this effect has significant implications on determination of planar flame speeds (and Markstein lengths as well) and, as sch, will be important to consider in spherical flame experiments. Figre 2b shows the nstretched flame speeds calclated throgh extrapolation to zero stretch over ranges of data that start at different radii and span 1 cm. For example, the extrapolated brning velocity for a starting radis of.5 cm is calclated by linear regression of flame speed vs. stretch data over the radis range from.5 to 1.5 cm, the extrapolated brning velocity for a starting radis of 1. cm is calclated over the radis range from 1. to 2. cm, and so on. In performing data redction over ranges that inclde only flame radii less than half of the chamber radis, the extrapolated flame speed varies by 1%. This implies a significant limitation in the crrent methods sed for constant-pressre cylindrical bombs and indicates the need for more restrictive data redction and/or correction factors, or alternate experimental designs for high pressre flame speed determination. 5

6 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames 3. Analysis of the Flow Field Effect To demonstrate the sensitivity of flame speed measrement to flow field deviation, we will derive a relation between the relative deviation from the ideal flow field, δ = Δ/, and the relative deviation in instantaneos flame speed, ε = s /s. The rate of propagation of a premixed flame throgh a moving gas can be expressed as the sm of two components: the flame speed, s, relative to the gas in the direction normal to the flame front and the gas velocity,, namely, dr f dt = s + (4) where the gas velocity has been taken in the direction normal to flame propagation for simplicity of presentation. Eqation (4) gives the propagation rate as a fnction of the actal flame speed and gas speed. However, calclation of s nder the assmption of an ideal flow field will yield an erroneos qantity for nbrned flame speed, s,apparent, which is related to the apparent gas velocity,,apparent, throgh, apparent ( 1) s, apparent = σ (5) by (1), where σ = ρ /ρ b is the nbrned to brned gas density ratio. Therefore, these assmptions yield the expression, drf = s, apparent +, apparent (6) dt By eqating the two expressions for dr f /dt, (4) and (6), s, = s + (7) apparent +, apparent and expanding for small deviation from the assmed flow field, Δ, sing (5) =, apparent + Δ =, apparent + ( σ 1) s δ (8) we can relate the deviation in the flow field, δ, to the deviation in the instantaneos flame speed, ε, throgh (7) and (8) s, apparent s ε = = ( σ 1) δ s where the gas velocity deviation, δ, can fond as a fnction of location for a given experiment sing diagnostics or approximate flow field soltions. In sitations where the deviation is small, sch that the strength and spatial distribtion of the hot gas expansion remains nearly constant, the flow velocity deviation, δ, is decopled from the flame propagation process and is a merely a fnction of position relative to the bondary conditions. Ths, along the coordinate of dr f /dt measrement, ξ, one can analytically or nmerically solve δ= δ(ξ), in which case the flow field deviation calclation for a given geometry has applicability for any mixtre. It is important to note from Eqation (9) that an O(1/(σ-1)) deviation in the flow field will yield an O(1) error in the apparent instantaneos flame speed, which indicates more serios errors are incrred for mixtres with higher density ratios. (9) 6

7 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames In physical terms, the otwardly propagating flame is a predominantly convective-driven phenomenon. Therefore, the qantity actally measred in the experiment, dr f /dt, is (σ-1) times more sensitive to nbrned gas velocity than the desired qantity, s, which is extracted from the dr f /dt data. In extracting flame speeds from flame propagation rate data, care mst be taken in general bt especially with mixtres with large density ratios, e.g. near stoichiometric mixtres. We will now formlate a simple model for the propagation of an otwardly propagating flame into an infinitely-long cylindrical pipe to illstrate the behavior of a propagating flame in a nonspherical chamber and develop an approximate model for the constant-pressre cylindrical bomb apparats. The flow field model presented herein tilizes potential flow to simlate the nbrned gas flow ahead of the flame prodced by hot gas expansion with appropriate cylindrical pipe bondary conditions. While it shold be noted that the actal flow is certainly not incompressible, a potential flow soltion represents a simple way of recreating the flow prodced by expansion of the combstion prodcts in a given geometry. The following soltion for the flow field is valid only for the nbrned mixtre, i.e. r > r f. Using soltions to Laplace s eqation, 2 φ = (15) the flow potential that satisfies the desired bondary conditions can be prodced throgh a sperposition of soltions. For small distortions of the flame, the flow created in the nbrned gas by expansion of the brned gas can be represented as a sorce at the origin φ = m(t ) center (16) 4 π ρ where m(t) is the time-varying strength and ρ is the radis from the origin. The strength of the sorce, m(t), is eqal to the amont of mass displaced by thermal expansion per nit time, 2 m( t) = ( σ 1) 4π r f s. (17) The flame propagation of an nconfined spherical flame is then given by the vector form of (4) where dr dt f = s n + (18) = φ. (19) The above relations sing the flow potential given by (16) form a soltion for a flame propagating into an nconfined environment. The apparent instantaneos flame speed is then calclated from dr f /dt sing the relations (18), (19), and (1). The model reqires vales for nstretched laminar flame speed, s o, Markstein length, L b, and density ratio, σ, as inpt parameters. To create the cylindrical pipe bondary, we can sperpose a ring of sorces with strength distribtion, q(t), at distance, η, from the origin smmed over all angles, γ, from to 2π, φ ring 2π = q( t) {[ x η cos( γ )] + [ y η sin( γ )] + z } 1/ 2 dγ (2) 7

8 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames where the strength distribtion, q(t), and distance, η, are chosen to satisfy r r= Rw φ = r cylinder r= Rw where r is the radial direction in the pipe and R w is the radis of the wall, and cylinder ring center = (21) φ = φ + φ. (22) Likewise, the nbrned gas velocity,, can be calclated by sing the cylindrical flow potential of (22) in eqation (19) and the apparent instantaneos flame speed, s,apparent, in a specific direction can be calclated sing (18) and (1). The model is applicable when the flow deviation is small, sch that the strength and spatial distribtion of the hot gas expansion remains nearly constant, and the flow velocity deviation, δ, is decopled from the flame propagation process and is a merely a fnction of flame position relative to the bondary conditions. Figre 3a illstrates the effect of the cylindrical pipe on the apparent flame speed of a mixtre with s o = cm/s, L =.212 cm, and σ = 5.36, and R w = 5 cm along the radial direction of propagation, which is the direction measred in cylindrical bombs. Of corse, the tre flame speed is nchanged, bt determination of the flame speed from dr f /dt, sing the typical data redction procedre for spherical flames described earlier, yields erroneos reslts for a flow field which departs from or assmptions. Flame speed (cm/s) Unconfined Model Cylindrical Model (Rw =5cm) Stretch (1/s) (a) Radial direction, r (cm) Axial direction, z (cm) (b) Figre 3: (a) Apparent instantaneos flame speed of nconfined and cylindrical models; (b) Flame shape evoltion of nconfined (dotted) and cylindrical (solid) models The plots reveal similarity of the nconfined and cylindrical cases for small radii (large stretch) and then a lower apparent flame speed for the cylindrical case for large radii (small stretch) de to a redced nbrned gas velocity in the radial direction. The linear region at small radii is ths the data range over which extrapolation shold be performed. Figre 3b displays the spatial evoltion of the flame front for the nconfined and cylindrical pipe models, from the perspective perpendiclar to the axial direction. The plot shows that the flame shapes for the two cases are similar for small radii, bt then differ significantly for large radii. The cylindrical pipe compresses the wave front in the radial direction and stretches it in the axial direction forming an elliptical shape. In the perspective along the axial direction (the view 8

9 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames observed in schlieren pictres of the cylindrical-bomb experiments), the flame front for the cylindrical confinement is circlar bt progresses at a lower rate than that of the nconfined case. It is expected that the flow in the cylindrical pipe may serve as an approximate model of the flow in a constant-pressre cylindrical bomb. Figre 4 compares the apparent flame speeds of an experiment of a hydrogen-air mixtre of eqivalence ratio 3. at 1 atm to the cylindrical pipe model sing s o o = cm/s, L =.212 cm, and σ = 5.36 as before (where the vales for s and L had been obtained from redction of the experimental data over.5 cm to 1.5 cm). The agreement between the experiment and the cylindrical pipe model is qite good despite its simplicity. The apparent flame speed of both cases agree with the nconfined case for small radii (large stretch), diverge from the nconfined case at moderate radii (moderate stretch), and decrease rapidly at large radii (small stretch). Flame speed (cm/s) Unconfined Model Cylindrical Model (Rw=5 cm) Cylindrical Experiment (Rw=5 cm) Stretch (1/s) Figre 4: Comparison of cylindrical model to cylindrical experimental data The reslt demonstrates the significance of flow field effects even in a constant-pressre combstion process. This marks the first time that the wall interference mentioned previosly in the literatre for cylindrical-bomb measrements has been qalitatively and qantitatively explained. The reslts from this analysis assist in the choice of data range and sggest possible correction factors for cylindrical bomb measrements. 4. Reslts and Discssion Laminar flame speeds for syngas mixtres of varios compositions p to 2 atm were measred sing more restrictive measres for data redction motivated by the reslts of the previos section. The syngas mixtres were composed of α CO2 CO 2 + (1-α CO2 )α H2 H 2 + (1-α CO2 )(1-α H2 )CO, where α CO2 = X CO2 /(X CO2 +X CO +X H2 ) is the CO 2 molar fraction in the fel mixtre, α H2 = X H2 /(X H2 +X CO ) is the H 2 -CO ratio, and X is the mole fraction. The oxidizer mixtres were composed of O N 2 for air and O 2 +7He for the oxygen-helim oxidizer. In light of the reslts from the flow field analysis section and previos stdies of ignition effects [2,26], the data range sed for extrapolation in this paper was.6 < r < 1.5 cm (=.3R w ), in order to minimize the effect of ignition, transient processes and non-niform flow field. Dring this range for the conditions stdied here, the flame front image was observed to be relatively smooth 9

10 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames and circlar and the stretched flame speed was observed to obey a linear dependence with stretch, as demonstrated in Figre 5a. Since flame speed measrements at low stretch rates are necessary for accrate extrapolation to zero stretch [39], the limitations imposed are expected to increase the ncertainty in the extrapolation compared to experiments that can accrately measre flame speeds at lower stretch rates. Laminar brning velocities of an H 2 -CO = 5:5 syngas mixtre in air at 1 atm are presented from the present experiment in Figre 6, along with experiments by McLean et al. [2], Hassan et al.[3], and Sn et al. [6], and simlations sing the recent mechanisms of Li et al. [15], Davis et al. [16], and Sn et al [6]. The measred brning velocities reach a peak vale of 186 cm/s at an eqivalence ratio of 2.. Since the present experiments and those of Sn et al. were both condcted in constant-pressre cylindrical bombs, the difference in planar brning velocities reslts primarily from different choices of the data range sed for extrapolation. For comparison, extrapolated planar brning velocities are presented in Figre 5b for flame speed-stretch data extrapolated over.6 cm < r <.3R w and for data extrapolated over.6 cm < r <.54R w (measrements were not made for flame radii larger than 2.7 cm in the experiment) along with the data of Sn et al. The corrected data range (.6 cm < r <.3R w ) is seen to raise the extrapolated brning velocity by as mch as 6% from the brning velocity fond from extrapolation over a wider range (.6 cm < r <.54R w ), which is more strongly inflenced by flow field deviations. While the deviation in the instantaneos flame speed is a fnction of density ratio and flame position only, the deviation in the extrapolated brning velocity is more complicated. The error in extrapolation also depends on Σ(κ i -κ mean ) 2 and κ mean, where κ i is the stretch rate at the i th point and κ mean is the mean stretch rate, as discssed in [39]. While Sn et al. did not report the data range sed for these measrements specifically, they referred to Tse et al. [7] for a detailed description of the experiment, where the range r <.61R w was sed. The data of Sn et al. and the data extrapolated over.6 cm < r <.54R w nearly coincide for eqivalence ratio less than 2., bt the data of Sn et al. are still lower at rich conditions. The difference in the vales obtained sing the two extrapolation ranges of the present experimental data demonstrates the significance of restricting data redction to small radii in constant-pressre cylindrical bombs. Flame speed (cm/s) phi=4. phi=.6 phi=1. phi= Stretch (1/s) (a) Brning velocity (cm/s) cm to.3rw.6 cm to.54rw Sn et al Eqivalence ratio Figre 5: (a) Extrapolation of flame speeds to zero stretch rate sing.6 cm < r <.3R w and (b) comparison of laminar brning velocities sing different data ranges for extrapolation for an H 2 - CO = 5:5 syngas mixtre in air at 1 atm (b) 1

11 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames As Figre 6 shows, while the experiments and simlations are consistent for lean mixtres, the vales diverge as the mixtre becomes richer. Despite the fact that all for data sets were from stretch-corrected measrements of spherical flames, no two data sets agree well over the whole range of eqivalence ratio and some differ by p to 4%. While Hassan et al. and McLean et al. report ncertainties (with 95% confidence) of 12% and 6%, respectively, these vales differ by 4% at very rich conditions. The disagreement in the experimental data sggests that the ncertainties for flame speed measrements are mch higher than previosly sggested. Given the sbstantial variation in the data, significantly more data and more rigoros calclations of ncertainty for these experiments are necessary before qantitative conclsions can be gathered and the performance of these mechanisms can be properly assessed. Brning velocity (cm/s) Present Work McLean et al. Sn et al. Hassan et al. Li et al. Davis et al. Sn et al Eqivalence ratio Figre 6: Laminar brning velocities for an H 2 -CO = 5:5 syngas mixtre in air at 1 atm Laminar brning velocities over a range of eqivalence ratios of an H 2 -CO = 25:75 syngas mixtre in oxygen-helim oxidizer for 1 and 2 atm are presented in Figre 7a. The measred brning velocities reach peak vales of 16 cm/s at an eqivalence ratio of 1.8 and 81 cm/s at an eqivalence ratio of 1.6, for 1 and 2 atm, respectively. The present measrements are higher than those of Sn et al. at 1 atm by more than 1% at lean and stoichiometric conditions. They agree for most conditions at 2 atm, except for rich conditions where the present measrements are lower by more than 1%. Simlations sing Li et al. and Sn et al. yield vales within 5% of the present high-pressre measrements, except for rich conditions, where the present measrements indicate a stronger adverse dependence of flame speed with eqivalence ratio than predicted by Sn et al. and Davis et al. at 2 atm and Li et al. at 1 and 2 atm. 11

12 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames Brning velocity (cm/s) atm 1 atm Eqivalence Ratio Brning velocity (cm/s) atm, phi= Molar CO2 fraction in fel (a) (b) Figre 7: Laminar brning velocities at elevated pressres of an H 2 -CO = 25:75 syngas mixtre in an O 2 :He = 1:7 oxidizer varying (a) eqivalence ratio and (b) CO 2 concentration in the fel mixtre for eqivalence ratio = 1. from present experiments (solid symbols), experiments by Sn et al. (open symbols), and simlations sing Li et al. (bold lines), Davis et al. (solid lines), and Sn et al. (dashed lines) Laminar brning velocities over a range of CO 2 concentrations of an H 2 -CO = 25:75 syngas mixtre in oxygen-helim oxidizer are presented in Figre 7b for 1 atm at stoichiometric conditions. The measred brning velocities indicate a decreasing linear dependence of brning velocity with CO 2 diltion. The experimental data for nstretched laminar brning velocity agree well with planar calclations sing the three mechanisms for each CO 2 fraction as well as the decreasing linear slope in flame speed with increasing CO 2 diltion. The three mechanisms yield similar reslts, bt Sn et al. predicts slightly higher flame speeds than Li et al., which predicts slightly higher flame speeds for low CO 2 concentrations and a slightly stronger adverse dependence of CO 2 concentration on the flame speed than Davis et al. 5. Conclding Remarks The effect of flow field deviation de to constant-pressre non-spherical chambers is investigated experimentally and analytically, revealing that the reslting deviation in the measred instantaneos flame speed is (σ-1) times the deviation in the flow field, where σ is the nbrned to brned gas density ratio. Since the density ratio for typical hydrocarbon-air mixtres is ~5-9, the flame speed measrement is extremely sensitive to departres from the theoretical flow field. A simple model is formlated to demonstrate the behavior observed in constant-pressre cylindrical bombs. In cylindrical chambers, where the flow is typically most constrained in the plane of measrement (radial direction), failre to consider this effect reslts in lower vales for the measred flame speed. We plan to contine this work to qantify this effect more accrately for the constant-pressre cylindrical bomb with detailed simlations. Planar brning velocities were measred for H 2 /CO/CO 2 mixtres varying in eqivalence ratio from.6 to 4., pressre from 1 to 2 atm, and CO 2 diltion from to 25%. The data range sed for extrapolation to zero stretch was chosen to be.6 cm < r < 1.5 cm (=.3R w ), where R w is the chamber wall radis, motivated by the reslts of the flow field analysis. The corrected data range (.6 cm < r <.3R w ) is seen to raise the extrapolated brning velocity by as mch as 6% from the brning velocity fond from extrapolation over a wider range (.6 cm < r <.54R w ), 12

13 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames which is more strongly inflenced by flow field deviations. While the experimental data and predictions for brning velocity agree reasonably well at lean conditions, large discrepancies occr at rich conditions. Comparison of the available stretch-corrected flame speed data for the conditions stdied reveals sbstantial variation, indicating that significantly more data and more rigoros calclations of ncertainty are necessary before qantitative conclsions can be made and the performance of kinetic mechanisms can be properly assessed. Acknowledgments The research at Princeton University was spported by the U. S. Department of Energy by grant #DE-FG26-6NT42716 and the Petrolem Research Fnd from American Chemistry Society by grant PRF# 4346-AC5. We wish to thank Andrew Kelley of C.K. Law s grop for helpfl discssions and se of his flame front detection program and Marcos Chaos for his help with simlations and writing this paper. References [1] N. Peters, Symposim (International) on Combstion 21 (1986) [2] I.C. McLean, D.B. Smith and S.C. Taylor, Symposim (International) on Combstion 25 (1994) [3] M.I. Hassan, K.T. Ang, G.M. Faeth, Jornal of Proplsion and Power 13 (1997) [4] C. M. Vagelopolos, F. N. Egolfopolos, Symposim (International) on Combstion, 25 (1994) [5] J. Natarajan, S. Nandla, T. Liewen, J. Seitzman, ASME Trbo Expo 25, Reno-Tahoe, Nevada, Jne 6-9, 25. [6] H. Y. Sn, S. I. Yang, G. Jomaas, C. K. Law, Proceedings of the Combstion Institte 31 (27) [7] S. D. Tse, D. L. Zh, C. K. Law, Proceedings of the Combstion Institte 28 (2) [8] X. Qin, Y. J, Proceedings of the Combstion Institte 3 (25) [9] G. Rozenchan, D.L. Zh, C.K. Law, S.D. Tse, Proceedings of the Combstion Institte 29 (22) [1] S. Kwon, L.K. Tseng, G.M. Faeth, Combstion and Flame 9 (1992) [11] D. Bradley, R.A. Hicks, C.G. Lawes, G.W. Sheppard, R. Wooley, Combstion and Flame 115 (1998) [12] R.A. Strenlow, L.D. Savage, Combstion and Flame 31 (1978) [13] S. Gordon and B.J. McBride, NASA Report No. SP-273 (1971). [14] R.J. Kee, J. F. Grcar, M.D. Smooke, J. A. Miller, Sandia Report SAND , [15] J. Li, A. Kazakov, Z. Zhao, M. Chaos, F.L. Dryer, J.J. Scire, International Jornal of Chemical Kinetics 39 (27) [16] S.G. Davis, A. Joshi, H. Wang, and F.N. Egolfopolos, Proceedings of the Combstion Institte 3 (25) [17] G.H. Markstein, Non-Steady Flame Propagation. Pergamon, New York, 1964, p. 22. [18] S.C. Taylor, Brning Velocity and Stretch, Ph.D. thesis, University of Leeds, Leeds UK, [19] D.R. Dowdy, D.B. Smith, S.C. Taylor, A. Williams, Symposim (International) on Combstion 23 (199) [2] D. Bradley, P.H. Gaskell, X.J.G, Combstion and Flame 14 (1996) [21] V.F. Karpov, A.N. Lipatnikov, P. Wolanski, Combstion and Flame 19 (1996) [22] P. Clavin, Progress in Energy and Combstion Science 11 (1985) [23] L. Qiao, C.H. Kim, G.M. Faeth, Combstion and Flame 143 (25) [24] J.H. Tien, M. Matalon, Combstion and Flame 84 (1991) [25] L.K. Tseng, M.A. Ismail, G.M. Faeth, Combstion and Flame 95 (1993) [26] S.Y. Liao, D.M. Jiang, J. Gao, Z.H. Hang, Q. Cheng, Fel 83 (24) [27] B. Lewis, G.J. Von Elbe, Jornal of Chemical Physics 2 (1934) [28] D. Bradley, A. Mitcheson, Combstion and Flame 26 (1976) [29] P.G. Hill, J. Hang, Combstion Science and Technology 6 (1988) [3] D.A. Daly, J.M. Simmie, J. Würmel, N. Djebaili-Chameix, C.E. Paillard, Combstion and Flame 125 (21)

14 5 th US Combstion Meeting Paper # A16 Topic: Laminar Flames [31] K. Saeed, C.R. Stone, Combstion and Flame 139 (24) [32] K. Takizawa, A. Takahashi, K. Tokhashi, S. Kondo, A. Sekiya, Combstion and Flame 141 (25) [33] F. Parsinejad, C. Arcari, H. Metghalchi, Combstion Science and Technology 178 (26) [34] Z. Chen, Y. J, 45th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, Jan. 8-11, 27. [35] J.K. Bechtold, M. Matalon, Combstion and Flame 67 (1987) [36] O.C. Kwon, G. Rozenchan C.K. Law, Proceedings of the Combstion Institte 29 (22) [37] C.K. Law, G. Jomaas, and J.K. Bechtold, Proceedings of the Combstion Institte 3 (24) [38] K.T. Ang, M.I. Hassan, G.M. Faeth, Combstion and Flame 19 (1997) [39] Z. Zhao, A. Kazakov, J. Li, F.L. Dryer, Combstion Science and Technology 176 (24)