[American Institute of Aeronautics and Astronautics 36th AIAA Aerospace Sciences Meeting and Exhibit...

Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

Simultaneous CH Planar Laser-Induced Fluorescence andParticle Imaging Velocimetry in Turbulent Flames

Jeffrey M. Donbar and James F. DriscollDepartment of Aerospace Engineering, University of Michigan

Ann Arbor, MI 48109

Campbell D. CarterInnovative Scientific Solutions, Inc.

Dayton, OH 45440

AbstractThe present work addresses the question ofwhether the chemical reaction zone within aturbulent, high-Reynolds-number jet flame isthin—and can be modeled using strained wrin-kled laminar flamelet theory—or is thick—andmust be modeled using distributed reaction zonetheory. The region near the wrinkled instanta-neous stoichiometric contour is identified usingCH Planar Laser-Induced Fluorescence (PLIF)imaging and the strain on this interface is meas-ured using simultaneous Particle Imaging Velo-cimetry (PIV) diagnostics. With a separate setof simultaneous images, the relative position ofthe CH and OH radicals is observed using PLIF.It is found that the CH reaction zone remainsthin and rarely exceeds 1 mm, even near theflame tip in a high-Reynolds-number (18,600)jet flame. The mean thickness of the CH reac-tion layer (5CH) increases from 0.3 to 0.8 mmin the streamwise direction; this is expected be-cause the scalar dissipation rate is known to de-crease with downstream distance and shouldcause a corresponding increase in 8CH. However,8CH also increases with jet velocity, which is notpredicted by theory. Furthermore, the meanstrain rates on the stoichiometric contour in-crease in the streamwise direction, which is con-trary to previous predictions. Thus, strain ratedoes not, in general, scale with the local dissipa-tion rate in a turbulent flame, and this aspect ofthe counterflow flow analogy is not valid. It isconcluded that unsteady laminar flamelet con-cepts are consistent with most of the presentobservations, but a method independent of the

local dissipation rate is needed to predict thelocal strain rate.

IntroductionOften, the analogy is made between seg-

ments of turbulent nonpremixed jet flames and asteady laminar counterflow diffusion flame thatis selected to have the same strain rate as theturbulent flame segment; particularly, this anal-ogy is used in "laminar flamelet models" [1,2].The flamelet regime, by definition, exists overregions in which the flame front is thinner thanthe smallest scale of turbulence, and thus theflame front can be modeled as an ensemble oflaminar flamelets. For the flamelet analogy tobe realistic, several factors must be known:1. if the reaction layers are sufficiently thin

(rather than thick, distributed zones);2. if the reaction-layer properties (e.g. radical

concentrations, thickness) scale as predictedby counterflow theory;

3. how frequently the layers extinguish ormerge; and

4. if the strain on the turbulent reaction layersis proportional to the dissipation rate as incounterflow flames [2].

If all of these conditions are satisfied, it becomespossible to apply several promising modelingideas which draw heavily on flamelet theory,including the original flamelet model [2], theLocal Integral Moment (LIM) model [3,4], theLarge Eddy Simulation (LES) model [5], and Di-rect Numerical Simulation (DNS) [6]. To date,lack of information on the structure of turbulent


reaction zones has prevented validation of thesemodels.

Most of what is known regarding reaction-zone structure comes from Planar Laser-InducedFluorescence (PLIF) imaging of OH in turbulenthydrogen-air flames [7-9]. While these imageshave been revealing, there is some questionabout the use of OH as a flame front marker.Because of the relatively slow removal of OH bythree-body collisions [10], OH can be convecteddownstream, making it difficult to determine theprimary fuel reaction zone. Consequently, OHis primarily a marker of the radical recombina-tion zone and does not necessarily identify theflame front. CH and C2 are better markers of theprimary reaction zone, since they are createdand removed by two-body reactions during pri-mary fuel decomposition.

Previous attempts at CH imaging in non-premixed flames have been hampered by lowconcentrations (tens of ppm or less as opposedto 103 - 104 ppm for OH) and interfering fluo-rescence from Polycyclic Aromatic Hydrocar-bons (PAHs), a known soot precursor.Consequently, the few studies involving planarimaging of CH have been limited to premixedflames [11-15] or regions of nonpremixedflames which contain no precursors, e.g., thebase region of lifted flames [16-19]. Starner etal. [20] and Barlow et al. [21], who obtained CHimages up to 30 jet diameters downstream in apiloted flame near extinction, show the onlyimages, to the authors' knowledge, of CH in thefar field of a nonpremixed flame. Overall, theCH images shown in the aforementioned studiesare similar in nature; they all indicate that theCH is within a thin, often highly contorted layerlocated slightly to the rich side of stoichiomet-ric.

While no attempts have been made in jetflames, researchers have used Particle ImagingVelocimetry (PIV) to determine strain andstretch in premixed flames. Driscoll et al. [22]determined aerodynamic strain rate, rate offlame stretch, vorticity, and shear strain rateduring a flame-vortex interaction. Similarly,Mueller et al. [23] studied repeatable flame-vortex interactions from the wrinkled flameletto local extinction regimes and generated a timehistory of flame stretch, strain, and vorticity.Nye et al. [24] measured flame curvature andstrain rate along a V-flame interacting with aKarman vortex street. All of these studies wereconducted in laminar flames, while PIV in tur-

bulent reacting flows, either premixed or non-premixed, has been used primarily within thepast few years [25,26].

In order to record strain rate, conditioned onthe reaction zone, simultaneous PIV and PLIFimages are required, the feasibility of which hasonly recently been demonstrated. Frank et al.[27] used PIV and PLIF imaging of biacetyl (areactant marker) in an unsteady premixed flameto identify velocity vectors relative to scalarcontours of fuel fraction. Rehm and Clemens[28] employed PIV and OH PLIF in turbulentplanar nonpremixed H2/N2-air flames, whileHasselbrink et al. [29] investigated a transversemethane jet flame in cross flow; however, noattempt was made in either case to compute lo-cal reaction-zone strain rates due to the breadthof the OH zones. Prior to the present work, noattempts have been made to perform CH fluo-rescence simultaneously with PIV.

Experimental ApparatusThe axisymmetric burner consisted of a cen-

tral fuel tube surrounded by a coflowing oxidizerstream. The 5-mm-diameter tube had a lipthickness of 1.47 mm and was positioned ap-proximately 15 mm downstream of the exitplane of the 150-mm-diameter coflow. For allexperiments, the coflow velocity was set to 0.3m/s. Flow rates of fuel, diluent, and oxidizerwere metered with Tylan mass flow controllers.The burner was mounted on a three-dimensionaltraversing system so that measurements could betaken at various positions from flame base totip.

Imaging of CH in nonpremixed flames isdifficult for the reasons alluded to earlier. Inorder to alleviate the PAH-interference prob-lem, nonsooting permanently blue nonpremixedjet flames, where the fuel is heavily diluted (~70-80% by volume) with N2 and the oxidizer ispure O2, were used in this study. Two methaneflames were studied extensively, correspondingto jet Reynolds numbers of 9,100 and 18,600;pertinent parameters for each flame are listed inTable 1. Aside from improving the feasibility ofimaging the CH layer, additional benefits to us-ing permanently blue flames have been realizedas well. The pure oxygen coflow producesflames substantially shorter than methane-airflames (60 vs. 200 jet diameters); the resultingflames are in the momentum dominated flowregime, i.e., buoyancy effects are negligible.


Decker and Yamazaki [30] have developed anondimensional parameter, £, based on theRichardson number (Ri), the ratio of buoyantforce to momentum flux, in order to quantifythe effects of buoyancy. They suggest that aflame is momentum dominated provided that £< 5; as Table 1 indicates both flames are wellbelow this level. The increase in stoichiometricmixture fraction, Zst, also pulls the flame frontradially inward to a point just inside the shearlayer; here, the turbulence intensity is muchhigher than found in typical methane-air flames.Lastly, the stability of the permanently blueflames permits them to remain attached to thetube lip even at exit velocities approaching 70m/s.

The experimental details have been de-scribed elsewhere [31] and will only be brieflyreviewed. A Nd:YAG-pumped dye laser wastuned to the Q,(7.5) transition of the B2Z~-X2n(V = 0, v" = 0) band (X = 390.30 nm) of CH andfluorescence from the A-X (1,1), (0,0) and B-X(0,1) bands (X = 420 - 440 nm) was recorded.This approach provides sufficient separationbetween pump and detection wavelengths to en-sure adequate rejection of scattered incidentlight, with the added advantage of increased fluo-rescence yield over schemes where the A-X(0,0) band is pumped and the A-X (0,1) (k =489 nm) band is detected. The 230-nm-thicksheet was passed through the center of the jet,and the resulting images collected using a fastNoct-Nikkor 58-mm/fl.2 lens and focused ontoa Princeton Instruments (PI) intensified chargecoupled device (ICCD) array. In order to im-prove the weak fluorescence signal, the ICCDpixels were binned 2 X 2 , resulting in an effec-tive spatial resolution of 122 Jim/pixel. It wasimperative to reject the strong scattering fromthe PIV particles to ensure the quality of the CHimages; this was accomplished through the use ofSchott color glass filters KV-418 and BG-1 [31].Note that for OH PLIF, a second Nd:YAG-pumped dye laser was tuned to the R^S.5) tran-sition of the A2E+-X2n (1,0) band (X -281.34nm) and fluorescence from the A-X (1,1) and(0,0) bands (A, = 306 - 320 nm) was collectedwith a UV-Nikkor 105-mm/f4.5 lens and imagedonto a second PI ICCD array (located on theopposite side of the burner from the CH ICCD).Temporal separation of the CH and OH probelasers was set to 0.5 (isec.

A standard two-color scheme was used forthe PIV; residual "red" dye laser output (616

nm) used in producing the CH excitation beamwas used as the red pulse and the 532-nm outputof the second Nd:YAG provided the green pulse.Prior to entering the test section, these twosheets were overlapped with the CH excitationsheet. Time separation between pulses wasmonitored with a photodiode and digital oscillo-scope and set to 5 and 10 \is for the Re =18,600 and 9,100 flames, respectively. Particlescattering from each sheet was collected onto asingle high-resolution (2036 by 3060 pixels,each 9 nm square) Kodak DCS 460 color digitalcamera; advantages of digital PIV and the cross-correlation scheme used to process the particleimages are discussed in detail by Gogineni et al.[32]. Velocity vectors were resolved using 128-pixel-square interrogation regions, correspond-ing to 1.5-mm-square regions of physical space.The interrogation regions were 75% overlappedsuch that the velocity vectors are spaced every0.375 mm on a regular grid. Overall, the frac-tion of good vectors was typically 90-95%; badvectors were removed and replaced by an inter-polated value from surrounding good vectors.The 75% overlap provided good vectors in closeproximity to each bad vector and thus reducedthe interpolation error. Nominal 0.5-|J.m A12O3particles were introduced in both the coflow andfuel tubes at such a level to ensure sufficient seeddensity while at the same time minimizing inter-ferences with the CH PLIF images.

Results

Reaction Zone Structure and ThicknessFor the purpose of studying reaction zone

structure, CH PLIF was performed without PIVand its accompanying seed particles. Two viewswere obtained at each location for all fourflames studied. The side-view orientation in-cludes the standard vertically oriented laser sheetpassing through flame center with the ICCD inposition to collect fluorescence at 90°. Alterna-tively, the 45° view is obtained by rotating thelaser sheet 45° while also rotating the camera45° in the same direction, in order to maintainthe 90°-collection geometry; by this means a"cross-sectional" view of the flame surface isobtained (the cross section of a cylindrical flamesurface would be an ellipse).

The cross-sectional view provides informa-tion on potential out of plane curvature in the


side view as well as insight on the degree of CHlayer wrinkling. A set of 45°-view images takenat x/Lf> st = 0.33 in a Re = 18,600 flame areshown in Fig. 1; the field of view for all imagesis 19.8 X 27.6 mm. By this location, the tur-bulence level has created a condition in whichthe general elliptical nature of the flame baseregion has been replaced by a wide variety ofshapes and sizes. The distribution is evenly splitbetween small structures with sharp curves andstructures with larger radius of curvature. Pre-dominantly the regions are thin, and thicker re-gions can be a result of the layer becomingaligned with the laser sheet.

The side-view CH images contain both left-and right-hand flame segments, with a field ofview equal to 24 X 36 mm. A set of side-viewimages from x/L( St = 0.67 is provided in Fig. 2.By this location, the connected reaction zone ofFig. 1 has become "pinched" or fragmented intoseveral pieces. Usually, the "pinching" createsseparate upper and lower burning surfaces withvoid between. The degree of wrinkling and pres-ence of extremely sharp curves is obvious ineach of the images. A potential cause for the"pinching" and fragmenting will be discussed inthe CH PLIF/PIV section.

By acquiring several hundred side-view im-ages of the CH layer only, the layer thicknesswas quantified. Figure 3 shows the mean CHlayer thickness (FWHM), 5CH, vs. nondimen-sional flame height for a variety of jet Reynoldsnumbers. Thickness was determined by taking ahorizontal cut through the image and countingconsecutive pixels above a threshold corre-sponding to half the maximum intensity.Hence, thickness is not necessarily measured in adirection normal to the CH surface, potentiallyresulting in an inflated value. Also, the finiteresolution of the imaging system might lead tosome overestimate of 6CH, particularly on sur-faces with a high degree of wrinkling (which ismore pronounced in the Re = 18,600 flame).Nonetheless, 8CH < 1 mm, even at the tip of theRe = 18,600 flame. Furthermore, two othertrends are distinctly noticeable: 1) the increasein 8CH with axial location for all flames and 2)the increase in 5CH due to increased jet velocity.The former can be attributed to the decrease inscalar dissipation from base to tip [33-36]; inconjunction with the conversion from a top-hatto Gaussian profile, local gradients must decreaseas well, thereby widening any isocontour of mix-ture fraction. However, the increase in 8CH with

jet-exit velocity is not predicted by flamelettheory. One possible explanation is that thesmall-scale structures at higher Reynolds num-bers are thickening the CH layer, i.e., it is nottruly flamelet-like, or layer thickening may be aresult of unsteady strain effects. Note that thereported CH thickness is similar to the value of0.6 mm measured in a nonpremixed laminarmethane-air flame [37].

CH PLIF/PIV ResultsA sample CH PLIF image with overlaid ve-

locity vectors is shown in Fig. 4. The image wasobtained at x/Lf St = 0.67 in a Re = 18,600flame. The concomitant large stoichiometricmixture fraction, Zst = 0.56, and heat release actto retard the velocity decay such that the meancenterline velocity is ~ 55 m/s. In order to visu-alize the vortical structures, 75% of the meancenterline velocity has been subtracted from allvectors, and thus vectors in the low-speed co-flow side are pointing downward. Several fea-tures of velocity-field/CH-layer interactions arenoticeable, particularly the large structure de-forming the top right and the small vortex dis-torting the middle of the right-hand side of theflame. The upper left-hand side shows a rela-tively weak CH region that is being acted uponby two small vortical structures, one from thejet fluid and the other from the oxidizer. Therelative position of the reaction zone and theshear layer is evidenced by the close proximityof downward-pointing vectors to the CH layer.

Figure 5 is also a CH PLIF/PIV image from*/Lf, st = 0-67 in a Re = 18,600 flame; however,it shows a pinched CH layer and the velocityfield associated with that phenomenon. Onceagain, 75% of the mean centerline velocity hasbeen subtracted from all of the vectors. It ap-pears as though a pair of large-scale structureshas merged in order to split the reaction zone.Evidence of a counter-clockwise-rotating struc-ture (associated with the left-hand side) is visiblein the central portion of the jet while portionsof a clockwise-rotating vortex can be seen in thelower-right portion of the image. Mungal et al.[38] have observed large-scale structures merg-ing to induce quasi-periodic burn out of theflame tip in jet flames; however, the velocity oftheir structures is roughly half the speed of thestructures measure herein.

With a simultaneous CH PLIF image andplanar velocity field, determination of the strainrate on the stoichiometric contour is possible


(and it will be shown that the CH layer lies closeto the stoichiometric contour). For an axisym-metric configuration, the flame-surface strainrate is given by Mueller et al. [23]:

du.dr

r , *dx dr

(1)

Here, r and x are the radial and axial coordinates,respectively, while u is the velocity and n is theflame-front normal. After alignment of the PIVand PLIF fields of view, coordinates of the CHlayer center were extracted from the CH imageat 0.375-mm increments, and each discrete CHflame-front point was matched with the nearestPIV velocity (from the regular PIV grid). Thus,a maximum spatial displacement error in eitherthe axial or radial direction was half the gridspacing or 0.1875 mm. Straight line segmentswere used to connect these points, and from thatline slope a corresponding perpendicular lineslope was determined. Using this slope, a nor-mal vector to the wrinkled contour (pointingtoward the oxidizer side) was determined. All ofthe partial derivatives in Eq. (1) were computedusing standard central differences on the regu-larly spaced velocity grid. This procedure wasrepeated at each CH coordinate within an image,typically 75 -100, to determine instantaneousstrain rate along the entire flame front.

A sub-region of a simultaneous PIV/CH PLIFimage taken at x/L{ St = 1 in a Re = 18,600 flameis shown in Fig. 6. The strong vortex has pulledthe flame tip over, turning it almost 180° fromthe original flow direction. Velocity spacing andsubtraction are the same as Figs. 4 and 5; hencethe structure shown is convecting at roughly 13m/s. The effect on the CH layer is most notice-able at the top, inner surface of the flame andleft-hand side of the vortex where the signallevel is noticeably weaker than the surroundinglayer. It is also interesting to note that the im-age shown in Fig. 6 is similar to computed im-ages of a single vortex interacting with a verticalflame sheet [39].

The instantaneous strain rate on the CHlayer (along the left-hand side of Fig. 6) was de-termined using Eq. (1) and is shown in Fig. 7 vs.the tangential flame coordinate (S); also shownis the axial velocity along the CH contour. Inthe regions of low CH PLIF signal, the strainrate is on the order of 15,000 s"1, constitutingsevere extensive strain, yet the flame does not

extinguish (this is in agreement with the coun-terflow calculations to be discussed and shown inFig. 10). This high strain rate is a result of thelarge change in axial velocity, over a small dis-tance, coupled with the dramatic change inflame orientation. Conversely, large compres-sive strain rates (-5,000 s"1) along the bottomcoincide with thicker CH layers. Some cautionshould be taken when looking at such high strainrates based on PIV measurements, as a velocityerror of ±1 m/s (out of 60 m/s) between adjacentregions can result in an error of ±2700 s"1 instrain. Regardless, the measurements indicateregions of large extensive and compressivestrain within a turbulent flame without flameextinction. It should also be noted that theseare instantaneous snapshots and one cannot de-termine how long these structures have acted orwill act on the flame surface. Some indicationof the rapid velocity change associated with thevortex can be noted in Fig. 7, as the axial ve-locity drops from 50 to 20 m/s along the flamesurface. Especially notable are the large drops(5-8 m/s) in the region of S = 0.9 which corre-spond to regions of high strain.

The statistics of the surface-averaged strainrate are quite revealing. Figure 8 shows the av-erage strain rate on the CH layer; ten images ateach of the four measurement locations wereanalyzed using Eq. (1), yielding 1000 individualstrain measurement points. The Re = 9,100flame shows a relatively constant average strainon the CH layer of 150 s"1; this is in stark con-trast to the Re = 18,600 flame which displays anincrease in average strain with downstream dis-tance. The disparity between the two cases re-sults because the Re = 9,100 flame has a morelaminar-like CH structure while the CH layers inthe Re = 18,600 flame are more highly wrinkledand contorted. While the averages are all posi-tive (extensive strain), regions of negative strainand corresponding thicker CH layers are alsoobserved (see lower left portion of Fig. 6 ).However, the extensive strain tends to domi-nate, especially downstream.

It was previously noted that the far-fieldstrain rate, according to theory, should scale asx'2. This theory, however, assumes the samedegree of wrinkling everywhere within theflame, which is not observed. The presentfindings indicate that the local degree of wrin-kling can be a dominant factor in determininglocal strain rate. Further proof of this is thatthe final term in Eq. (1) is dominant near the


flame tip. There, the flame orientation ischanging rapidly (nx and nr vary greatly), and inaddition the surface is adjacent to the shear layerwhere cHiy^/- is substantial. Clearly, modelsshould predict the degree of wrinkling to repro-duce a realistic strain rate along the reactionzone.

The CH/PIV data allows assessment of therelationship between mixture fraction and ve-locity. The issue is thoroughly discussed by Kuo[40] and is only briefly reviewed herein. In alaminar jet flame, the ^-momentum equationand the fuel species conservation, equation writ-ten in terms of mixture fraction, have identicalsolutions, provided diffusivities are equal andexit profiles of velocity and mixture fraction aretop-hat in nature. With exact similarity be-tween the processes of mass and momentumtransfer, the velocity at the stoichiometric sur-face, Us, can be written as Us = ZSt • U0, whereU0 is the jet exit velocity. For each of themeasurement locations in both the Re = 9,100and Re = 18,600 flames, Us (axial velocity only,as radial velocities are low) was computed andplotted in Fig. 9; the 1-a deviations for eachpoint are indicated for each point. Figure 9 in-dicates that the CH layer (an isocontour of mix-ture fraction) stabilizes along an isocontour ofaxial velocity, approximately 24 and 44 m/s, forthe Re = 9,100 and Re = 18,600 flames, respec-tively. Both of these values are slightly higherthan expected from the above equation (21 and37 m/s). Regardless, even in turbulent flames itappears as though the mixture fraction and ve-locity profiles are similar.

Computations and CH/OH ImagingFor purposes of comparison, a series of

computations using steady counterflow flameswas conducted. The "fuel" side consisted of afuel-product mixture at Z = 0.8, where ZSt =0.56 (same as Re = 18,600 flame), and the oxi-dizer was pure O2 at 300 K (same as the coflow).Figure 10 shows the computed CH-layer thick-ness (FWHM) and distance from peak CH molefraction to ZSt as a function of both strain rateand the scalar dissipation rate at thestoichiometric contour (%s). As the dissipationrate increases, the computed thickness decreasesin agreement with the experimental data in Fig.3. The disparity between measured and calcu-lated 5CH (calculated value is smaller) may be aresult of the concerns stated in the section(Reaction Zone Structure and Thickness) on ex-

perimental determination of 8CH- As for thedistance between the stoichiometric surface andpeak CH level, the separation at low strain ratesis ~ 1.5 mm; however, as the strain rates in-crease the profiles narrow, and the separationdistance at 2,000 s"1 is less than 0.5 mm. Thus,the strain rates measured along the maximumCH contour are representative of the strain onthe stoichiometric contour.

In order to quantify the differences betweenCH and OH imaging, a set of simultaneous im-ages was acquired. Figure 11 shows the combinedCH/OH images throughout the entire length of aRe = 18,600 permanently blue jet flame. Thesix uncorrelated images were taken at each ofthe four measurement locations, xfLf St = 0, 1/3,2/3 (c & d), and 1 (e & f). The center of theimages corresponds to x = 10, 80, 160, and 200mm from the jet exit. In 11 a), both layers arestraight and vertically oriented with an occa-sional "bump" that appears to result from astructure within the jet fluid. Proceeding to 11b), these "bumps" have increased in magnitude;the CH layer remains thin while the OH layerhas thickened and shows some signs of being in-fluenced by outer edge effects. The most dra-matic changes occur in going from 11 b) to c).Here, the CH layer becomes highly distortedwith sharp cusps; furthermore, the two CH flamesurfaces nearly coalesce at the top of the image.A thin band of high OH signal is adjacent to theCH, but the overall OH-layer width has increasedsignificantly and shows distinct internal struc-ture. In lid), taken at the same axial locationas c), the CH reaction layers have "pinched"together resulting in separate upper and lowerreaction zones with OH filling the void betweenthem. In turn, this "pinching" phenomenonleads to structures like those appearing in 11 e);here, small pockets of OH are surrounded bythin CH layers within the burning jet cone. It istheorized that these islands are partially pre-mixed pockets that are created during the"pinching" process. Finally, Fig. 11 f) corre-sponds to the instantaneous stoichiometricflame tip, Lf St; as expected, significant quanti-ties of OH persist downstream in the hot-gasregion.

Figure 11 illustrates the discrepancy inthickness between CH and OH regions; espe-cially downstream, the CH layer remains thinand flamelet-like while the OH layer becomesseveral cm thick, more like a distributed reactionzone. In contrast to the findings of several


studies of lifted hydrocarbon and diluted hydro-gen flames [16, 34, 41], no "holes" are seen atthe base of the flame; in lifted flames, vorticalstructures grow within the mixing layer resultingin flame-surface holes [16]. As the flames inthis study are attached, and perhaps owing to theincreased stability of permanently blue flames[42], this phenomenon is not observed. Fur-thermore, in the far field there is an abundanceof OH radicals, making it difficult to extinguishthe flame front, regardless of strain rate.

Conclusions1. The CH layer remains thin and rarely ex-

ceeds 1 mm, even near the flame tip in ahigh-Reynolds-number (18, 600) jet flame.The CH layer thickness increases in thestreamwise direction; this is expected as thescalar dissipation rate has been shown to de-crease in the downstream direction of jets.However, SCH also increases with jet veloc-ity, which is not predicted by theory.

2. The CH layer in turbulent flames is highlycontorted and wrinkled, but rarely broken.In the far field, some "pinching" and frag-menting of the CH layer does occur. It ap-pears as though merging of large-scalestructures is responsible for the splinteringof the reaction zone further downstream.Clear evidence of the velocity field acting todistort the CH layer is noted.

3. The strain rates on the stoichiometric con-tour are small near the flame base and in-crease in the streamwise direction, contraryto model predictions. The base region doesnot experience the maximum strain rates, aspredicted, because the flame tends to alignwith the mean flow near the base. Near thetip the flame experiences large extensivestrain rates up to 20,000 s'l where theflame experiences large curvature. Thestrain on the wrinkled stoichiometric con-tour does not scale in the same manner asthe predicted strain rate on the mean flamecontour, and thus it is important for modelsto simulate the wrinkled flame shape in or-der to account for effects of strain.

4. On the average, axial velocity along theflame surface is shown to be constant overthe entire flame length. This indicates ve-locity and mixture fraction profiles in tur-bulent flames are similar as predicted inlaminar flames.

5. Simultaneous CH/OH imaging clearly dem-onstrates that these radicals mark differentregions within the flame. The CH layer re-mains thin and flamelet-like along the entireflame and provides an indication of the pri-mary hydrocarbon consumption zone. Onthe other hand, OH is thicker and exhibitsdefinite structure, distributed-like, while pro-viding a better marker of the radical recom-bination zone.

AcknowledgmentsThis work has been supported by the Air ForceResearch Laboratory, Propulsion Directorate,Wright-Patterson Air Force Base, OH, underAFOSR Contract Nos. F33615-92-C-2202 andF33615-97-C-2702, Dr. Julian Tishkoff, Con-tract Monitor. JMD has received support fromthe USAF Palace Knight Program.

References1. Peters, N., "Laminar Flamelet Concepts in

Turbulent Combustion," Twenty-First Sym-posium (International) on Combustion,The Combustion Institute, Pittsburgh, PA,1986, pp. 1231-1250.

2. Peters, N., and Williams, F.A., "LiftoffCharacteristics of Turbulent Jet DiffusionFlames," AIAA J. 21:421-429 (1983).

3. Suresh, N.C., Dahm, W.J.A., and Tryggva-son, G., "LIM Modeling of Chemical Reac-tions in Spatially and TemporallyDeveloping Shear Flows", AIAA Paper 94-0870, Reno, NV, 1994.

4. Tryggvason, G., and Dahm, W.J.A., "AnIntegral Method for Mixing, Chemical Re-actions, and Extinction in UnsteadyStrained Diffusion Layers," Combust.Flame 83:207-220, 1991.

5. Grinstein, F.F., and Kailasanath, K.,"Three-Dimensional Numerical Simulationsof Unsteady Reactive Square Jets," Com-bust. Flame 100:2-10, 1995.

6. Mahalingam, S., Chen, J.H., and Vervisch,L., "Finite Rate Chemistry and TransientEffects in Direct numerical Simulations ofTurbulent Nonpremixed Flames," Combust.Flame 102:285-297,1995.

7. Seitzman, J.M., Ungiit, A., Paul, P.H., andHanson, R.K., "Imaging and Characteriza-tion of OH Structures in a Turbulent Non-premixed Flame," Twenty-Third


Symposium (International) on Combus-tion, The Combustion Institute, Pittsburgh,PA, 1990, pp. 637-644.

8. Johnson, A.W., Sreenivasan, K.R., andWinter, M., "The Thickness Distributionof OH Regions in a Turbulent DiffusionFlame," Combust. Sci. Tech. 89:1-7, 1993.

9. Clemens, N.T., Paul, P.H., and Mungal,M.G., "The Structure of OH Fields in HighReynolds Number Turbulent DiffusionFlames," accepted for publication: Com-bust. Sci. Tech.

10. Barlow, R.S., Dibble, R.W., Chen, J.-Y., andLucht, R.P., "Effect of Damkohler Numberon Superequilibrium OH Concentration inTurbulent Nonpremixed Jet Flames," Com-bust. Flame 82:235-251, 1990.

11. Chen, Y.-C., and Mansour, M.S.,"Simultaneous Rayleigh Scattering and La-ser-Induced CH Fluorescence for ReactionZone Imaging in High-Speed Premixed Hy-drocarbon Flames," Appl. Phys. B 64:599-605, 1997.

12. Nguyen, Q.-V., and Paul, P.H., "The TimeEvolution of a Vortex-Flame InteractionObserved via Planar Imaging of OH andCH," Twenty-Sixth Symposium(International) on Combustion, The Com-bustion Institute, Pittsburgh, PA, 1996, pp.357-364.

13. Paul, P.H., and Dec, J.E., "Imaging of Re-action Zones in Hydrocarbon-Air Flamesby Use of Planar Laser-Induced Fluores-cence of CH," Opt. Lett. 19:998-1000,1994.

14. Tsujishita, M., Ipponmatsu, M., and Hi-rano, A., "Visualization of the CH Mole-cule by Exciting C2Z+ (v = 1) State inTurbulent Flames by Planar Laser-InducedFluorescence," Jpn. J. Appl. Phys.32:5554-5569, 1993.

15. Alien, M.G., Howe, R.D., and Hanson,R.K., "Digital Imaging of Reaction Zonesin Hydrocarbon-Air Flames Using PlanarLaser-Induced Fluorescence of CH and C2,"Opt. Lett. 11:126-128, 1986.

16. Schefer, R.W., Namazian, M., Filtopoulos,E.E.J., and Kelly, J., "Temporal Evolutionof Turbulence/Chemistry Interactions inLifted, Turbulent-Jet Flames," Twenty-FifthSymposium (International) on Combus-tion, The Combustion Institute, Pittsburgh,PA, 1994, pp. 1223-1231.

17. Schefer, R.W., Namazian, M., and Kelly, J.,"Stabilization of Lifted Turbulent-JetFlames," Combust. Flame 99:75-86, 1994.

18. Schefer, R.W., Namazian, M., and Kelly, J.,"CH, OH, and CH4 Concentration Meas-urements in a Lifted Turbulent-Jet Flame,"Twenty-Third Symposium (International)on Combustion, The Combustion Institute,Pittsburgh, PA, 1990, pp. 669-676.

19. Namazian, M., Kelly, J.T., and Schefer,R.W., "Near-field Instantaneous Flame andFuel Concentration Structures," Twenty-Second Symposium (International) onCombustion, The Combustion Institute,Pittsburgh, PA, 1988, pp. 627-634.

20. Starner, S.H., Bilger, R.W., Dibble, R.W.,Barlow, R.S., Fourguette, D.C., and Long,M.B., "Joint Planar CH and OH LIF Imag-ing in Piloted Turbulent Jet DiffusionFlames Near Extinction," Twenty-FourthSymposium (International) on Combus-tion, The Combustion Institute, Pittsburgh,PA, 1992, pp. 341-349.

21. Barlow, R.S., Dibble, R.W., Starner, S.H.,Bilger, R.W., Fourguette, D.C., and Long,M.B., "Reaction Zone Structure in DiluteMethane Jet Flames Near Extinction,"AIAA Paper 90-0732, Reno, NV, 1990.

22. Driscoll, J.F., Sutkus, D.J., Roberts, W. L.,Post, M.E., and Goss, L.P. "The Strain Ex-erted by a Vortex on a Flame—Determinedfrom Velocity Field Images," Combust. Sci.and Tech. 96:213-229, 1994.

23. Mueller, C.J., Driscoll, J.F., Reuss, D.L., andDrake, M.C., "Effects of Unsteady Stretchon the Strength of a Freely-PropagatingFlame Wrinkled by a Vortex," Twenty-SixthSymposium (International) on Combus-tion, The Combustion Institute, Pittsburgh,PA, 1996, pp. 347-355.

24. Nye, D.A., Lee, J.G., Lee, T.-W. and San-tavicca, D.A., "Flame Stretch Measure-ments During the Interaction of PremixedFlames and Karman Vortex Streets Using,PIV," Combust. Flame 105:167-179,1996.

25. Mungal, M.G., Lourenco, L.M., andKrothapalli, A., "Instantaneous VelocityMeasurements in Laminar and TurbulentPremixed Flames Using On-Line PIV,"Combust. Sci. Tech. 106:239, 1995.

26. Muniz, L., and Mungal, M.G.,"Instantaneous Flame-Stabilization Veloci-


ties in Lifted-Jet Diffusion Flames," Com-bust. Flame 111:16-31, 1997.

27. Frank, J.H., Lyons, K.M., and Long, M. B.,"Simultaneous Scalar/Velocity Field Meas-urements in Turbulent Gas-Phase Flows,"Combust. Flame 107:1-12, 1996.

28. Rehm, J.E., and Clemens, N.T., "APIV/PLIF Investigation of Turbulent Pla-nar Non-Premixed Flames," AIAA Paper97-0250, Reno, NV, 1997.

29. Hasselbrink, E.F., Mungal, M.G., and Han-son, R.K., "Planar Velocity Measurementsand OH Imaging in a Transverse Jet,"AIAA Paper 97-0118, Reno, NV, 1997.

30. Becker, H.A., and Yamazaki, S.,"Entrainment, Momentum Flux and Tem-perature in Vertical Free Turbulent Diffu-sion Flames," Combust. Flame 33:123-149, 1978.

31. Carter, C.D., Donbar, J.M., and Driscoll,J.F., "Simultaneous CH Planar Laser-Induced Fluorescence and Particle ImagingVelocimetry in Turbulent NonpremixedFlames,"^/?/?. Phys. B, 66:129-132 (1998).

32. Gogineni, S., Goss, L., Pestian, D., andRivir, R., "Two-Color Digital PIV Em-ploying a Single CCD Camera," Acceptedfor publication: Expts. in Fluids.

33. Nandula, S.P., Brown, T.M., and Pitz,R.W., "Measurements of Scalar Dissipationin the Reaction Zones of Turbulent Non-premixed H2-Air Flames," Combust. Flame99:775-783, 1995.

34. Effelsberg, E., and Peters, N., "Scalar Dis-sipation Rates in Turbulent Jets and Jet Dif-fusion Flames," Twenty-Second Symposium(International) on Combustion, The Com-bustion Institute, Pittsburgh, PA, 1988, pp.693-700.

35. Everest, D.A., Driscoll, J.F., Dahm, W.J.A.,and Feikema, D.A., "Images of the Tem-perature Field and Temperature Gradientsto Quantify Mixing Rates within a Non-Premixed Turbulent Jet Flame," Combust.Flame 101:59-68, 1995.

36. Feikema, D.A., Everest, D.A., and Driscoll,J.F., "Images of Dissipation Layers toQuantify Mixing Within a Turbulent Jet,"AIAA J. 34:2531-2538, 1996.

37. Norton, T.S., and Smyth, K.C., "Laser-Induced Fluorescence of CH in a LaminarCH4/Air Diffusion Flame: Implications forDiagnostic Measurements and Analysis of

Chemical Rates," Combust. Sci. Tech. 76:1-20, 1991.

38. Mungal, M.G., Karasso, P.S., and Lozano,A., "The Visible Structure of Turbulent JetDiffusion Flames: Large-Scale Organiza-tion and Flame Tip Oscillation," Combust.Sci. and Tech. 76:165-185, 1991.

39. Ashurst, W.T., and Williams, F.A., "VortexModification of Diffusion Flamelets,"Twenty-Third Symposium (International)on Combustion, The Combustion Institute,Pittsburgh, PA, 1990, pp. 543-550.

40. Kuo, K.K., Principles of Combustion, JohnWiley and Sons, New York, 1986.

41. Chen, T.H., and Goss, L.P., "StatisticalOH-Zone Structures of Turbulent JetFlames from Liftoff to Near-Blowout,"Combust. Sci. Tech. 79:311-324, 1991

42. Du, J., and Axelbaum, R.L., "The Effectsof Flame Structure on Extinction of CH4-O2-N2 Diffusion Flames," Twenty-SixthSymposium (International) on Combus-tion, The Combustion Institute, Pittsburgh,PA, 1996, pp. 1137-1142.

TABLE 1. Characteristics of two flames studiedin this experiment. UQ is jet exit velocity, XCH4and XN2 are mole fractions of methane and ni-trogen in fuel mixture, Lf vis and Lf St are visibleand stoichiometric flame lengths, respectively,ZSt is stoichiometric mixture fraction, and £quantifies the effects of buoyancy.

Un (m/s)X TH4

X N2

Lfv i , (mm)L f s , (mm)zs,1 = Ri1/3-(x/d)

Re = 9,10031.2

0.1940.8064003000.672.50

Re= 18,60065.2

0.2970.7033202400.561.25


en

91= PA 91 = PA


f-d

>dHoTsOOVOoo

II<D

qm̂

Iorn

I0(Nm

q q—H' Om m

o•*'CO

IC5COm

~C5(Nm

Oo

q(N

oo

t/5(U

§Ku<D3C/5

O•—ta

p/x p/x


<U

cd

ooO

CNO

<u• • — iD

0)

03

(L)O,

D

+->o

•*—•

G

's

CJCO

ffiO(D

•*-*C

ooo

(ram) H3§

rno

OHHfe


34.0-

33.0 -

32.0-

31.0-

30.0-2.0 -1.0 0.0

r/d

1.0

FIGURE 4. Simultaneous CH PLIF/PIV image in a Re = 18, 600 flame, x/Lf s = 0.67.

Velocity vectors spaced every 375 (4,m on a regular grid and regions are 75% overlapped;

75% of the mean centerline velocity has been subtracted from all vectors.


34.0-

33.0

32.0

31.0

30.0-

f.f'f tl t ff ft t t f M

f f f t t t* f f f t t t

-2.0 -1.0 0.0

r/d

1.0 2.0

FIGURE 5. Simultaneous CH PLIF/PIV image in a Re = 18, 600 flame, x/Lf §t = 0.67.

Velocity vectors spaced every 375 |J,m on a regular grid and regions are 75% overlapped;

75% of the mean centerline velocity has been subtracted from all vectors.


42.0-

41.5 -

41.0-

-1.5 1.0

r/d

FIGURE 6. Sub-region of a simultaneous CH PLIF/PIV image in a Re = 18,600flame at jc/Lf St = 1. 75% of mean centerline velocity has been subtracted from allvectors.

en[/

w

-5,000

-10,000

60

50

40

30

20

10

D"

0.5 1 1.5 2TANGENTIAL FLAME COORDINATE S (cm)

2.5

FIGURE 7. Instantaneous flame strain rate, 8 , and axial velocity vs. flameo

tangential coordinate~i.e., the distance along the flame front—for the image shown

in Fig. 6. For reference, the enclosed squares are indicated on both Figs. 6 and 7.


2,000

1,500 -

1,000 -

500 -

0.8

FIGURE 8. Mean strain rate, e , measured along the instantaneous CH layer.

\J\J

55

50

45

40

35

30

25

7.O

i • i • i • i • i • ,

-

-

I

-

-

1 i 1 [

-

1 [

-

i--

U Re — 18,6OOA Re = 9,1OO

: } } { * :+ . 1 , 1 . 1 , 1 , 1

0.2 0.4 0.6x/L

0.8

FIGURE 9. Mean axial velocity measured along the instantaneous CH layer.All points show 1-s variation in velocity at flame front.

STOICHIOMETRIC SCALAR DISSIPATION RATE - xs (s'1)

E

X

100 1,000 10,0002.0

1.5 -

—© -CH THICKNESS

0.50 -

0.0100

T-I 0.50

- 0.40

- 0.30

- 0.20

- 0.10

0.01,000 10,000

STRAIN RATE - Es (s'1)

FIGURE 10. Calculated CH layer thickness and distance from peak CH mole fraction to stoichiometriccontour vs. strain rate, e , and stoichiometric scalar dissipation, % , in steady planar counterflow flames.s sFuel-product mixture at Z = 0.8 and pure oxygen as reactants.

Copyright © 1998, American Institute of Aeronautics and Astronautics, Inc.

0.0 I ' I ' I ' I ' I ' I r

-2.0 0.0 2.0 r/d

° 2 T l t l T ° 2CH4 & N2

FIGURE 11. Composite image of flame structure for a jet Reynolds number = 18,600permanently blue nonpremixed flame. OH intensity from low (blue) to high (red) withCH in black (thin layer inside of OH). The A/Lf St locations are: a) 0, b) 1/3, c) 2/3,d)2/3,e)l ,andf)l .

[American Institute of Aeronautics and Astronautics 36th AIAA Aerospace Sciences Meeting and Exhibit...

Documents

Transcript of [American Institute of Aeronautics and Astronautics 36th AIAA Aerospace Sciences Meeting and Exhibit...