CF Weigand Et Al

Combustion and Flame 144 (2006) 205224www.elsevier.com/locate/combustflame

Investigations of swirl flames in a gas turbinemodel combustor

I. Flow field, structures, temperature, andspecies distributions

P. Weigand, W. Meier , X.R. Duan 1, W. Stricker, M. AignerInstitut fr Verbrennungstechnik, Deutsches Zentrum fr Luft- und Raumfahrt (DLR), Pfaffenwaldring 38,

D-70569 Stuttgart, Germany

Received 22 November 2004; received in revised form 2 June 2005; accepted 8 July 2005

Available online 21 September 2005

Abstract

A gas turbine model combustor for swirling CH4/air diffusion flames at atmospheric pressure with good opticalaccess for detailed laser measurements is discussed. Three flames with thermal powers between 7.6 and 34.9 kWand overall equivalence ratios between 0.55 and 0.75 were investigated. These behave differently with respect tocombustion instabilities: Flame A burned stably, flame B exhibited pronounced thermoacoustic oscillations, andflame C, operated near the lean extinction limit, was subject to sudden liftoff with partial extinction and reanchor-ing. One aim of the studies was a detailed experimental characterization of flame behavior to better understand theunderlying physical and chemical processes leading to instabilities. The second goal of the work was the estab-lishment of a comprehensive database that can be used for validation and improvement of numerical combustionmodels. The flow field was measured by laser Doppler velocimetry, the flame structures were visualized by planarlaser-induced fluorescence (PLIF) of OH and CH radicals, and the major species concentrations, temperature, andmixture fraction were determined by laser Raman scattering. The flow fields of the three flames were quite sim-ilar, with high velocities in the region of the injected gases, a pronounced inner recirculation zone, and an outerrecirculation zone with low velocities. The flames were not attached to the fuel nozzle and thus were partially pre-mixed before ignition. The near field of the flames was characterized by fast mixing and considerable finite-ratechemistry effects. CH PLIF images revealed that the reaction zones were thin (0.5 mm) and strongly corrugatedand that the flame zones were short (h 50 mm). Despite the similar flow fields of the three flames, the oscillatingflame B was flatter and opened more widely than the others. In the current article, the flow field, structures, andmean and rms values of the temperature, mixture fraction, and species concentrations are discussed. Turbulenceintensities, mixing, heat release, and reaction progress are addressed. In a second article, the turbulencechemistryinteractions in the three flames are treated. 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

* Corresponding author. Fax: +49 711 6862 578.E-mail address: [email protected] (W. Meier).

1 Present address: Southwestern Institute of Physics, P.O. Box 432, 610041 Chengdu Sichuan, Peoples Republic of China.0010-2180/$ see front matter 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.doi:10.1016/j.combustflame.2005.07.010

n and

lence

Rayleigh scattering, laser Raman scattering, coher- In the study discussed here, a nozzle with two

ent anti-Stokes Raman scattering (CARS), or laser-induced fluorescence (LIF); species concentrations byLIF or Raman scattering; flame structures by planarLIF (PLIF) or PIV; and the mixture fraction by Ra-

concentric air swirlers and an annular fuel supply be-tween them was used for CH4/air diffusion flames,with thermal powers up to 35 kW at atmosphericpressure. The combustion chamber enabled almost206 P. Weigand et al. / Combustio

Keywords: Gas turbine; Model combustor; Swirl flame; Turbutechniques

1. Introduction

Swirl flames are used extensively in practical com-bustion systems because they enable high energy con-version in a small volume and exhibit good ignitionand stabilization behavior over a wide operating range[14]. In stationary gas turbine (GT) combustors, theyare used mostly as premixed or partially premixedflames, and in aero engines, as diffusion flames. Toreduce pollutant emissions, especially NOx , today theflames are operated generally very lean [57]. Un-der these conditions, the flames tend to exhibit unde-sired instabilities, e.g., in the form of unsteady flamestabilization or thermoacoustic oscillations. The un-derlying mechanisms of the instabilities are basedon the complex interaction between flow field, pres-sure, mixing, and chemical reactions, and are notwell enough understood to date. Detailed measure-ments in full-scale combustors are hardly possible,and very expensive and numerical tools have not yetreached a sufficient level of confidence to solve theproblems. A promising strategy lies therefore in theestablishment of a laboratory-scale standard com-bustor with practical relevance and detailed, com-prehensive measurements using nonintrusive tech-niques with high accuracy. The gained data set willbe used for validation and optimization of numeri-cal combustion simulation codes which then can beapplied to simulate the behavior of technical combus-tors. Intrusive probe measurements are less suited forthese applications as they disturb the local flow fieldand change the conditions for stabilization and forreactionlocally or even in general [8,9]. In turbu-lent reacting flows, the use of optical measurementtechniques is therefore essential for reliable infor-mation. Laser-based tools are the method of choiceoffering the potential to measure most of the impor-tant quantities with high temporal and spatial reso-lution, often as one- or two-dimensional images, andthe ability to perform the simultaneous detection ofseveral quantities [1013]. The flow field can be mea-sured by laser Doppler velocimetry (LDV) or particleimaging velocimetry (PIV); the temperature by laserman scattering.In recent years a variety of laser-based investiga-

tions in GT model combustors have been reportedFlame 144 (2006) 205224

chemistry interaction; Validation measurements; Laser

that, besides feasibility studies, concentrated on cer-tain aspects of the combustion process or modelvalidation. For example, Kaaling et al. [14] per-formed temperature measurements with CARS ina RQL (rich-quench-lean) combustor, and Kamp-mann et al. [15] used CARS simultaneously with 2-DRayleigh scattering to characterize the temperaturedistribution in a double-cone burner. In the same com-bustor, Dinkelacker et al. [16] studied the flame frontstructures with PLIF of OH and 2-D Rayleigh scat-tering. Fink et al. [17] investigated the influence ofpressure on the combustion process by applying PLIFof OH and NO in a LPP (lean prevaporized premixed)model combustor. With respect to NOx reductionstrategies, Cooper and Laurendeau [18,19] performedquantitative NO LIF measurements in a lean direct-injection spray flame at elevated pressures. Shih etal. [20] applied PLIF of OH and seeded acetone in alean premixed GT model combustor, and Deguchi etal. [21] used PLIF of OH and NO in a large practicalGT combustor. Hedman and Warren [22] used PLIFof OH, CARS, and LDV for the characterization ofa GT-like combustor fired with propane in order toachieve a better understanding of the fundamentalsof GT combustion. PLIF of OH was also applied byLee et al. [23] to study flame structures and instabil-ities in a lean premixed GT combustor, by Arnold etal. [24] to visualize flame fronts in a GT combustorflame of 400 kW, and by Fritz et al. [25] for revealingdetails of flashback. Lfstrm et al. [26] performeda feasibility study of two-photon LIF of CO and 2-Dtemperature mapping by LIF of seeded indium in alow-emission GT combustor. A comparison of twodifferent laser excitation schemes for major speciesconcentration measurements with laser Raman scat-tering was performed by Gittins et al. [27] in a GTcombustion simulator. At a high-pressure test rig ofthe DLR, various laser techniques (LDV, CARS, PLIFof OH and kerosene, and 2-D temperature imagingvia OH PLIF) were applied to GT combustors un-der technical operating conditions to achieve a betterunderstanding of combustor behavior and to validateCFD codes [2831].unrestricted optical access to the flames and was,thus, ideally suited for the application of laser mea-surement techniques. The velocity fields were mea-

n andP. Weigand et al. / Combustio

sured by 3-D LDV, the flame structures by PLIF ofOH and CH, and the joint probability density func-tions (PDFs) of temperature, major species concentra-tions, and mixture fraction by laser Raman scattering.Three flames with different characteristics were in-vestigated: flame A was operated at a specific powerrate of 42.4 MW/(m3 bar) that is comparable to thevalues of aeronautical or modern aero-derivative in-dustrial gas turbines, which are operated around 25to 70 MW/(m3 bar); flame B was chosen at a powerrate of 12.5 MW/(m3 bar) that is comparable to mostindustrial gas turbines, which are operated at 5 to20 MW/(m3 bar); and flame C was operated at thesame airflow as flame B but with reduced fuel sup-ply close to the lean extinction limit (with a powerrate of 9.2 MW/(m3 bar)). This is of interest becausemodern gas turbines in power plants are operated un-der extremely lean conditions to meet the emissionlimits. In addition, the flames were operated at threedifferent equivalence ratios to investigate the stabi-lization of the flames. Flame A with an equivalenceratio of = 0.65 burned stably, whereas flame B( = 0.75) emitted strong thermoacoustic noise, andflame C with = 0.55 operated close to the blowofflimit and randomly experienced sudden liftoff andreestablishment of stable operation.

The advantage of the combustor setup used wasthe excellent optical access to the flame zone, en-abling the collection of information from the wholearea around the flame zone in a burner that is closeto technical application. In particular, the detailedvelocity measurements at the nozzle exit result inwell-defined boundary conditions, which are impor-tant for numerical methods. One major goal of thework was the detailed experimental analysis of theflames to gain deeper insight into, e.g., the mixingand stabilization processes, the shape of the reactionzones and the regions of heat release, and effects ofturbulencechemistry interactions. The second goalwas the establishment of a comprehensive databasewhich can be used for the verification and improve-ment of combustion simulation codes. The presentarticle focuses on flow fields, on the distribution of thetemperature, the major species concentrations and themixture fractions, and on the instantaneous and meanflame structures. The turbulencechemistry interac-tions, which play an important role in these flames,are discussed in a second article [32]. The thermoa-coustic oscillations of flame B were analyzed previ-ously by phase-resolved measurements [3335]. Theresults from those investigations represent a supple-ment of the measurements without phase resolution

presented in the current article, and some of the find-ings are used here to support the discussion of thecharacteristics of flame B.Flame 144 (2006) 205224 207

2. Experimental

2.1. Combustor and flames

The gas turbine model combustor is schematicallyshown in Fig. 1. The burner was a modified versionof a practical gas turbine combustor with an air blastnozzle for liquid fuels [36]. Co-swirling dry air atroom temperature was supplied to the flame through acentral nozzle (diameter 15 mm) and an annular noz-zle (i.d. 17 mm, o.d. 25 mm contoured to an outerdiameter of 40 mm). Both air flows were fed froma common plenum with an inner diameter of 79 mmand a height of 65 mm. The radial swirlers consistedof 8 channels for the central nozzle and 12 chan-nels for the annular nozzle. The ratio of the air massflows through the annular and central nozzle was ap-proximately 1.5. Nonswirling CH4 was fed through72 channels (0.5 0.5 mm), forming a ring betweenthe air nozzles. Compared with an annular nozzlefor CH4 with a slit width of

n and

min.8.3.0

l mixrature

All three flamesestablished under globally lean 290 Hz. Flame C showed nearly the same frequency

conditionsshowed no soot production and burnedwith a light blue color. The flames appeared as type 2swirl flames [1,37] with a conically shaped toroidalflame zone at different opening angles and, as wasconfirmed by the velocity measurements, showed pro-nounced recirculation zones on the axis (inner recir-culation zone, irz) and near the walls of the com-bustion chamber (outer recirculation zone, orz). Thevisual appearance of the flames, despite the squarecombustion chamber, revealed good rotational sym-

spectrum as flame B but with a much reduced ampli-tude. Flame A emitted a rather broadband noise withsmall peak amplitudes at about 380 Hz.

2.2. Measuring techniques

All three velocity components were measured si-multaneously using commercial LDV systems (DISA/DANTEC) and a cw-Ar+ laser (Coherent, INNOVA90, operated at 1 W). The optical arrangement con-208 P. Weigand et al. / Combustio

Table 1Parameters of the three flames investigated

Air CH4sl/min g/min sl/min g/

A 850 1095 58.2 41B 218 281 17.2 12C 218 281 12.6 9

a Pth, thermal power; glob, equivalence ratio for the overaladiabatic temperature for the overall mixture with inlet tempe

to the flame. A conical top plate made of steel witha central exhaust tube (diam 40 mm, length 50 mm)formed the exhaust gas exit. The high velocity in theexhaust tube avoided any backflow from outside thecombustion chamber.

The three flames investigated were: flame A, withPth = 34.9 kW and an overall equivalence ratio ofglob = 0.65 that ran very stably; flame B, withPth = 10.3 kW and glob = 0.75, which exhibitedpronounced self-excited thermoacoustic oscillationsat a very high noise level; and flame C, operated closeto the lean extinction limit, with Pth = 7.6 kW andglob = 0.55, which randomly lifted off and rean-chored to the normal stabilization height. Table 1 listscharacteristic parameters of the investigated flames,i.e., the volume and mass flow rates of air and fuelas well as the resulting values for power and overallglobal values for equivalence ratio, mixture frac-tion, and adiabatic temperature (given the subscriptglob). The mass flows of the gases were controlledwith Brooks flow controllers (Type 5853S for air andType 5851S for CH4) with an accuracy of typically0.5%. As can be seen from Table 1, flames B andC had almost identical total flow rates but differentflame parameters. To achieve this, the air mass flowwas kept constant for both flames and only the fuelmass flow, which represents 7.3% of the total flow forflame B and 5.5% for flame C, was changed. This waschosen to achieve a high similarity of the velocity dis-tributions to exclude flow field effects as a source forthe different behavior of the two flames.metry.The swirl number S was calculated from the ve-

locity profile just above the nozzle exit neglecting theFlame 144 (2006) 205224

Ptha

(kW)glob f glob T glob ad

(K)34.9 0.65 0.037 175010.3 0.75 0.042 1915

7.6 0.55 0.031 1570

ture; f glob, mixture fraction for the overall mixture; T glob ad,T0 = 295 K.

pressure term according to

S = R

0 2uwr drR R

0 2u2r dr,

where u = axial velocity (m/s), w = circumferentialvelocity (m/s), = density (kg/m3), r = radius (m),and R = maximum radius of the nozzle exit (m). Theswirl numbers are S 0.9 for flame A and S 0.55for flames B and C. Given the fact that the nozzlewas contoured and a combustion chamber was usedwith an expansion factor D/d = 3.4 (D = diameterof combustion chamber, d = diameter of nozzle), vor-tex breakdown with establishment of an irz was to beexpected [38]. Due to the confinement, an orz wasalso found. The nozzle Reynolds number based on thecold inflow and the minimum outer nozzle diameter(25 mm) was about 15,000 for flames B and C andabout 58,000 for flame A.

All three flames did not burn directly at the fuelnozzle exit, but rather with a liftoff height of somemillimeters. In flame C, sudden liftoff (partial ex-tinction) and flashback randomly occurred (approx-imately 10 times per minute), as it was chosen tooperate close to the lean extinction limit that wasfound to be at = 0.53. The random liftoff whichreached up to a height of 3040 mm lasted typically100150 ms. Thus, flame C was, for about 2% of thetime, in the partially extinguished mode. Of the threeflames discussed here, flame B exhibited the highestnoise level with a quite discrete frequency of aboutsisted of a two-component system (DISA 55X,laser = 488 and 514.5 nm) and a single-componentsystem (DISA Flow Direction Adapter, laser =


514.5 nm), which were arranged orthogonally andboth used in the forward scatter mode. The laserbeams were transmitted via mono mode fibers intothe optical modules. A frequency shift of 40 MHz wasapplied for all three directions; the center frequencyof the detection was adapted to each measuring point.Focusing lenses with f = 300 mm were used; the re-sulting probe volumes were about 60 m in diameterand 1.0 mm in length for x and y directions, and120 m in diameter and 1.5 mm in length for the zdirection, corresponding to the axial (u), radial (v),and tangential (w) directions of the velocity, respec-tively. Because the spatial intensity distribution of theMie scattering with its maximum in forward directionis more than 100 times higher than at 90, the twosystems could be operated at the same wavelength(here 514.5 nm) because the forward-scattered sig-nals can be clearly discriminated by their intensities inthis orthogonal geometry. The detection optics werearranged in the yz plane at about 10 off axis andconsisted of commercial camera lenses (f = 85 mm,f/1.8 for x and y directions and f = 105 mm, f/4for the z direction) focusing the signals on photo-multiplier tubes. For the simultaneous detection andanalysis of the photomultiplier signals, three Dantecburst spectrum analyzers (BSA enhanced, 57N20 and57N35) were used with a record length of 64 points.ZrO2 particles with a diameter of approximately 2 mwere seeded as scatterers into the airflow. Becauseof the small size, the particles can follow flow fieldfluctuations up to a frequency of 1.2 kHz within anaccuracy of 99%. At each measuring location, typi-cally 10,000 to 15,000 validated velocity data wererecorded, except in the area of the flame zone, wheresometimes only 2000 samples were validated duringthe record time. For flames B and C, measurementsat lower heights were carried out coincidentally witha time filter of 2 s, thus providing also the Reynoldsstress tensors and cross moments. Using only the co-incidental values, the effective probe volume and,consequently, the data rate are drastically reduced,leading to much longer acquisition time. Therefore,this was done only at selected heights. For calcu-lating mean values, the noncoincidental values wereused for an improved statistic. The lowest height forLDV measurements was h = 1.5 mm for flame A. Inflames B and C an improvement of the setup enabledmeasurements as low as h = 1.0 mm. For simplicityin this study, these levels are all labeled h = 1 mm.

Planar laser-induced fluorescence (PLIF) of OHand CH radicals was applied to visualize the flamestructures. A Nd:YAG laser pumped optical paramet-ric oscillator (Spectra Physics GCR 290 and MOPO

730) was used to supply pulsed laser radiation for theexcitation of OH and CH radicals. The laser beamwas formed to a light sheet (h 45 mm) and irradi-Flame 144 (2006) 205224 209

ated vertically into the flame intersecting the flameaxis. The pulse energies were typically 3 mJ/pulsefor OH and 4 mJ/pulse for CH with a bandwidthof about 0.45 cm1. The sheet thickness was ap-proximately 0.25 mm in the imaged area. The re-sulting spectral laser intensities are on the order of16 MW/cm2 cm1 for CH and 12 MW/cm2 cm1for OH. Compared with the saturation intensities,which are around 1 MW/cm2 cm1 for the chosentransitions [39,40], the applied laser intensities arerelatively high and a significant degree of satura-tion is expected. The excited fluorescence was col-lected at 90 by a lens (for OH: achromatic UV lens,f = 100 mm, f/2, Halle Nachf.; for CH: camera lens,f = 50 mm, f/0.95, Canon) and, after spectral filter-ing, detected with an intensified CCD camera (forOH: LaVision Flamestar II; for CH: Roper Scientific).The laser pulse duration was 5 ns; the temporal detec-tion gate of the image intensifier was 50 ns for OHand 200 ns for CH (limited by irising effects of theimage intensifier). OH radicals were excited on theR1(8) line of the A2+X2 ( = 1, = 0) tran-sition at = 281.3 nm [41] and the fluorescence wasdetected through an interference filter in the wave-length region 312 10 nm. For CH, the Q1(7)line of the B2X2 ( = 0, = 0) band wasexcited at 390.3 nm [42,43]. For suppression of laser-scattered light and background radiation, a filter com-bination of a KV418 (Schott) and a short-pass filterwith a cutoff wavelength of 450 nm (Oriel) was usedin front of the camera, and only the fluorescence inthe BX (0,1) and AX bands around 430 nmwas detected. The A state is efficiently populated bycollision-induced electronic energy transfer from theB to the A electronic level [44,45].

For pointwise quantitative measurement of theconcentrations of major species (O2, N2, CH4, H2,CO, CO2, H2O) and temperature, laser Raman scat-tering was used [46]. The radiation of a flashlamp-pumped dye laser (Candela LFDL 20, wavelength = 489 nm, pulse energy Ep 3 J, pulse durationp 3 s) was focused into the combustion cham-ber, and the Raman scattering emitted from the mea-suring volume (length 0.6 mm, diam 0.6 mm)was collected by an achromatic lens (D = 80 mm,f = 160 mm) and relayed to the entrance slit of aspectrograph (SPEX 1802, f = 1 m, slit width 2 mm,dispersion 0.5 nm/mm). The dispersed and spatiallyseparated signals from the different species were de-tected by photomultiplier tubes in the exit plane of thespectrograph and sampled by boxcar integrators. Thespecies number densities were calculated from thesesignals using calibration measurements, and the tem-

perature was deduced from the total number densityvia the ideal gas law [46,47]. The simultaneous de-tection of all major species with each laser pulse also

n and

and C

max3. Results and discussion

3.1. LDV measurements

As expected for this type of confined swirl flame,the mean flow field of each of the three flames showsa strong inner recirculation zone (irz) along the axialcenterline as well as an outer recirculation zone (orz)near the walls of the combustion chamber, as can beseen in the vectorplots of the combined uv velocitiesdisplayed in Fig. 2. The inlet velocities at the lowestmeasuring height h = 1 mm correspond to the massflows, with the maximum values for the axial velocityof umax = 39.0 m/s in flame A, umax = 13.3 m/s inflame B, and umax = 13.0 m/s in flame C. The angleof the maximum mean velocity in the uv plane withrespect to the axial centerline is about 26 for all threeflames for h = 1020 mm. The measurements also

C (S 0.55). Nevertheless, the resulting flow fieldsare very similar, despite the different visible appear-ance of the three flames. The normalized mean axialvelocities (umean/umax) at h = 1 mm, illustrated inFig. 3, are almost identical and show that the inflowat this height extends radially from r = 5 to 15 mm.The isolines of umean = 0, plotted in Fig. 4, indi-cate the boundaries of the irz. It can be seen that forflames A and C, the irz reaches up to h 73 mm,whereas in flame B it ends at h 62 mm. In the nearfield of the nozzle, for h < 10 mm, the contours ofthe irz are nearly identical for all three flames, and inall three flames the irz extends below the lowest mea-suring level. It can therefore be assumed that the irzreaches even into the central air nozzle, which endsat h = 4.5 mm. In comparing the mean values ofthe different flames, it must, however, be consideredthat flame B is subject to periodic oscillations and that210 P. Weigand et al. / Combustio

Fig. 2. Vectorplots of combined uv velocities for flames A, B,the size of the combustion chamber.

enabled the determination of the instantaneous mix-ture fraction [48]. At each measuring location 500single-pulse measurements were performed within ascanning pattern of roughly 100 points, from whichthe joint probability density functions (PDFs) werecomputed. The choice of 500 samples turned out to bea good trade-off between measuring time and conver-gence of the mean and rms values. Studies in highlyturbulent regions of these flames revealed that the fi-nal values are reached to within 2% after 300 to 400samples. The measurement uncertainty for the meanvalues of temperature, mixture fraction, and molefraction of O2, H2O, and CO2 is typically 34%.revealed (not shown) that the ratio of the tangentialvelocity w and the axial velocity u is nearly doublefor flame A (S 0.9) compared with flames B andFlame 144 (2006) 205224

; negative u velocities are displayed in red. The lines indicate

Fig. 3. Radial profiles of the normalized mean axial velocity(u/u ) at h = 1 mm for flames A, B, and C.the time-averaged velocities represent an average notonly over turbulent fluctuations but also over periodicvariations (see discussion below).


Fig. 4. Isolines of umean = 0 representing the extension ofirz for flames A, B, and C.

Fig. 5 illustrates the mean values and rms fluctua-tions of the three velocity components of flame A forh = 1.5, 5, 15, and 45 mm. At h = 1.5 mm, the profileof the axial velocity u reflects the inflow of the freshgas at r 516 mm with maximum values around39 m/s and the irz with velocities of u 20 m/s.The radial velocity component v is negative for r >16 mm, reflecting the size of the orz. In the inflow,v is roughly half as large as u. The tangential ve-locity w is rather constant in the orz (w 10 m/s),and its radial profile displays two maxima in the re-gion of the inflow, which likely reflect the flows fromthe two air nozzles with the minimum between themoriginating from the fuel nozzle and its wake. Forr = 05 mm, w increases linearly with r , reflectingthe solid body rotation part of the vortex. The rmsvalues of u and v have a pronounced maximum inthe shear layer between the inflow and the irz, and vexhibits another maximum in the shear layer betweenthe inflow and the orz. The high level of the rms val-ues close to the flame axis demonstrates that the flowfield is subject to strong turbulent fluctuations in thisregion of the flame. At h = 5 mm, the gradients of theradial profiles of the mean and rms values have be-come smaller in comparison to h = 1.5 mm, but thebasic features of the flow are unchanged. In the orz,umean is close to zero but urms is 69 m/s. This showsthat u changes frequently its direction in the orz. At

h = 15 mm, the profiles have broadened and the re-verse flow on the axis reaches its highest negative ve-locity of umean 26 m/s. The orz has shrunk but isFlame 144 (2006) 205224 211

still discernible from the negative v component. Themean tangential velocity component indicates a solidbody vortex up to r 10 mm. For r > 10 mm, wmeandeclines but the shape of the radial profile does notresemble that of a potential vortex. wrms is quite con-stant over the radius, whereas urms and vrms exhibitbroad maxima. With increasing downstream position,the profiles smooth out and the orz vanishes, whereasthe irz reaches a radial expansion of r 13 mm ath = 45 mm.

In the shear layer between the inflow and the irz,large velocity fluctuations and the low mean veloc-ity generally cause a very high turbulence intensity(urms/umean 100%). Therefore, intense mixing ofthe cold fresh gas with hot burned gases coming fromthe irz can be expected in this region. Fig. 6 shows,for example, radial profiles of normalized urms ath = 10 mm, normalized by the maximum velocitiesat h = 1 mm; i.e., umax = 39.0, 13.3, and 13.0 m/sfor flames A, B, and C, respectively. The broad peaksof the urms values around r 6 mm indicate that theinstantaneous flow fields are subject to strong turbu-lent fluctuations and that the shear layer is not lo-cally stable. This finding is also supported by thesingle shot 2-D LIF images that are discussed in thefollowing paragraphs. In Fig. 6 it can also be seenthat for the three flames, the relative velocity fluctu-ations urms/umax are very similar and the values ofurms reach more than 50% of umax at r 310 mm.Therefore, the irz should not be regarded as a sta-ble structure with the fluid following streamlines thathardly vary their positions. However, from the mea-surements performed in this study, there was no indi-cation of coherent structures such as rotating vortexpairs.

To demonstrate that the average flow conditions atthe nozzle exit for flames B and C were similar, theradial velocity profiles of u, v, and w at the exit of thenozzle are plotted in Fig. 7. As clearly can be seen, themean profiles of all three velocity components matchvery well as intended, so that the mean flow field canbe excluded as the primary reason for the differentbehavior of these two flames. However, periodic vari-ations would not be revealed by the mean profiles.One also recognizes the good symmetry of the time-averaged velocity profiles, as is expected for a rota-tional symmetric flow. The dips in the profiles that canbe seen at r 1012 mm result from the wake of thefuel nozzle and have already disappeared at h = 5 mm(not shown) due to the high turbulence. The orz be-comes apparent by the inward-directed radial veloc-ity v in the region r > 15 mm. Further downstream,the different thermal powers and combustion temper-

atures of flames B and C lead to a different thermalexpansion which has an influence on the flow veloci-ties. To demonstrate this effect, Fig. 8 shows the radial

n andFig. 5. Radial profiles of the mean values (left side) and rms fluctuations (right side) of the three velocity components in flame Aat different heights.

profiles of u, v, and w at h = 30 mm. Here, the ax-ial velocity in flame B is significantly larger than inflame C; e.g., umax is 10.6 m/s in flame B and 8.3 m/sin flame C. The profiles of the radial velocity com-ponent are almost identical and those of the tangen-tial velocity show only slightly higher velocities forflame B. Thus, the different thermal expansion of theflames influences predominantly the axial velocity,

In flame B, LDV measurements have also beenperformed with phase resolution at heights h = 1and 5 mm. In those measurements, the phase ofthe acoustic oscillation was measured by a micro-phone simultaneously with two velocity components[34,35]. An important finding was that the irz andorz varied in size during an oscillation cycle, resem-bling a pumping motion: The irz varied mainly in212 P. Weigand et al. / Combustioas expected for confined flames. It is also typical ofconfined flames that the swirl number decreases withcombustion progress due to the axial acceleration.Flame 144 (2006) 205224the axial direction, and the orz in the radial direc-tion. Thus, although the mean flow fields of flames Band C look very similar, they are inherently differ-


Fig. 6. Radial profiles of urms at h = 10 mm for flames A,B, and C.

ent, with flame B exhibiting periodic variations super-posed onto turbulent fluctuations and flames A and Cexhibiting only turbulent fluctuations.

3.2. Flame structures from OH LIF and CH LIFmeasurements

In flames, OH can be found in detectable concen-trations at temperatures above approximately 1400 K,especially in fuel-lean mixtures [49]. The equilib-rium OH concentration increases exponentially withtemperature but differently for fuel-lean and fuel-richmixtures. Furthermore, OH is formed in superequi-librium concentrations in the reaction zones, and itsrelaxation to equilibrium by three-body collisions isquite slow at atmospheric pressure ( > 3 ms) [50].Thus, high OH LIF intensities can be regarded as anindicator of hot gas and/or reacting fuel/air mixtures.CH radicals are formed at high temperatures on thefuel-rich side of the reaction zone and have a muchshorter lifetime (10 ns) than OH radicals [51]. Thus,high CH concentrations can be interpreted as a markerfor the fuel consumption layer of the reaction zoneand, with some restrictions, as a qualitative measureof the heat release rate [52]. For illustration, Fig. 9shows the calculated profiles of the temperature andOH and CH mole fractions as a function of the mix-ture fraction f for a strained laminar CH4/air counter-flow diffusion flame with a strain rate of a = 400 s1[53,54]. Significant CH concentrations are presentonly in mixtures with f 0.050.08. In contrast, OHis also found in lean mixtures and covers a range off 0.0150.08. It can also be seen that CH is a fac-tor of about 500 lower in concentration than OH and,thus, much harder to detect. Although the turbulentflames investigated cannot be directly compared with

a counterflow diffusion flame, this example shows atleast qualitatively the characteristic behavior of OHand CH.Flame 144 (2006) 205224 213

With respect to the interpretation of the LIF im-ages presented here, one has to keep in mind thatthe LIF intensities are not necessarily proportionalto the species number density. The relative popula-tion of the initial state excited by the laser changeswith temperature (Boltzmann fraction f B). For OH,the Boltzmann fraction of the initial rotational state,N = 8, varies by less than 7% over the temperaturerange of interest (T 14002200 K). For CH, f B ofN = 7 decreases by roughly 20% over the tempera-ture range 1700 to 2200 K. Because the LIF signalsare quenching dominated, variations in quenching en-vironment can significantly influence the fluorescenceyield. An estimation was performed for OH using theLASKIN program [55] and the temperature and gascomposition from the strained laminar flame calcula-tion with a = 400 s1 already used in Fig. 9. It turnedout that the OH fluorescence yield varied by about15% over the range of interest. Taking into accountthat the composition of the flame under investigationmay deviate from the strained laminar flame compo-sition, the measured LIF signal intensities reflect theOH density roughly within 25%. For CH, the situationis more complex because quenching in two electronicstates, A and B, and predissociation in the B state playa role. However, in the thin layer of the reaction zonewhere CH is present, the gas composition and tem-perature do not change drastically and variations inquenching are expected to be small. Rensberger et al.[56] reported that changes in the fluorescence quan-tum yield after excitation of the B( = 0) state indifferent flames were small and that quenching var-ied by less than 50%. In the flames investigated here,variations should not be larger.

Fig. 10 shows typical OH single-shot LIF distri-butions for the three flames. The images display theregion r = 4141 mm and h 047 mm. For tem-peratures below 13001400 K, OH concentrationswere below the detection limit (dark areas). For allthree flames, the OH distributions cover broad areaswith strongly wrinkled contours and sometimes iso-lated regions (at least in a 2-D cut). These structuresyield a good impression of the turbulent transport andmixing processes within the flames. These imagesshow that the instantaneous flame structures (and verylikely also the instantaneous flow fields) are much lessuniform as might be assumed from the mean flowfield. The steep gradients of OH LIF intensities thatfrequently occurred may represent either a reactionzone or the boundary between cold and hot fluid. TheOH-free regions near the nozzle reflect the inlet flowof mostly unreacted fuel and air, which is directed di-agonally upward. The mean contours of these regions

can be better seen in the averaged images of Fig. 11(left). The highest mean OH LIF intensities and thus,within 25% uncertainty, the highest mean OH con-

n and214 P. Weigand et al. / Combustio

Fig. 7. Radial profiles of the simultaneously measured threevelocity components u, v, and w for flames B and C ath = 1 mm.

centrations are found in each flame in the irz shortlyabove the nozzle, with a maximum at h 10 mm.Here, flame C has the lowest mean OH concentration,followed by flame B with approximately 50% moreand flame A with approximately 75% more. Aroundh = 30 mm on the axis, the concentrations are ap-proximately three times less than at h = 10 mm foreach flame. These significant OH concentrations inthe irz in the averaged images (and also in most of

the single shot images) indicate a high temperaturein this region. The mixing of this hot gas from theirz with fresh gas from the nozzles presumably playsFlame 144 (2006) 205224

Fig. 8. Radial profiles of the simultaneously measured threevelocity components u, v, and w for flames B and C ath = 30 mm.

the key role in the ignition and stabilization of theflames. Investigations in flame B using planar two-line OH LIF thermometry showed that temperatureand OH concentrations were not generally well cor-related in that flame and that superequilibrium con-centrations contributed significantly to the high OHlevels within and some millimeters downstream ofthe reaction zones [57]. Comparison of the OH LIFdistributions from the three flames further shows that

flame C exhibits a smaller area containing OH, espe-cially in the orz. This is explained by the lower overalltemperature level of this flame (see Table 1) accord-


Fig. 9. Calculated profiles for temperature and concentra-tions of OH and CH radicals in a counterflow diffusion flamewith a strain rate of a = 400 s1.

ing to the global equivalence ratio. It is also seen thatfor flame A, the inlet flow of cold gas penetrates fur-ther downstream in comparison to the other flames.

The averaged CH LIF distributions displayed inFig. 11 (right) reflect the regions where flame reac-tions and heat release take place. The shapes of thethree flames are quite different: while flames A and Care conically shaped with opening angles /2 30and 45, respectively, flame B has a significantlylarger opening angle and is rather flat with /2 75.The difference between flames B and C is surprisingbecause their mean flow fields are quite similar andthe mean velocities are almost identical at h = 1 mm(see Fig. 7). Comparing the velocity fields and theCH LIF images it becomes apparent that the heat re-lease for flame C and especially for flame B does nottake place predominantly in the shear layer betweenthe irz and the inflow, as might be expected. The CHLIF distribution and the region of the shear layer arein good accordance only for flame A, whereas forflames B and C, the opening angles of the flame zonesare larger than that of the shear layer. The unusual be-havior of flame B is related to the periodic pulsationsand is addressed in Section 3.3. Further, it is impor-tant to note that for all three flames the regions ofheat release do not begin at the fuel nozzle, but ath 5 mm, h 4 mm, and h 6 mm for flames A,B, and C, respectively. Due to this liftoff, fuel, air,and exhaust gas are already partially premixed beforeignition. According to the CH LIF images, the heatrelease is complete at h 20 mm and h 40 mm forflames B and C, respectively. In flame A, there is stilla small amount of CH at the upper end of the mea-sured area at h = 47 mm.

For the interpretation of the averaged distributionsone has to keep in mind that flames B and C exhibitunsteady combustion behavior. Flame A is highly tur-

bulent but steady, and mean and rms values are re-lated to turbulent fluctuations. Flame C is subject tosudden partial extinction and can be regarded as bi-Flame 144 (2006) 205224 215

modal, i.e., either burning stably or not burning up toh = 3040 mm. However, the partially extinguishedstate is present only about 2% of the time and its con-tribution to the time-averaged values is small. Thus,flame C should be classified as nearly steady and itsfluctuations are caused predominantly by turbulence.For flame B, the situation is different, because thethermoacoustic pulsations are permanent and tempo-ral changes of the flame include turbulent fluctuationsas well as periodic variations. The mean species andtemperature distributions discussed in this article areaveraged over both turbulent and periodic changes. Todistinguish between them, phase-resolved measure-ments have to be performed that yield averaged valuesat distinct phase angles of the periodic pulsation. Suchmeasurements have also been performed for flame B,and the results are discussed in detail in separate pub-lications [3335].

The single-shot LIF distributions of CH, displayedin Fig. 12, show thin (0.30.5 mm) and strongly cor-rugated reaction zones which are, at least in the 2-Dcut, sometimes interrupted. Their shapes are domi-nated by the turbulent flow field and vary stronglyfrom shot to shot. Some samples exhibit only weakflame reactions, while others possess strongly dis-torted and intense reaction zones. Analysis of a largernumber of single shots revealed that in flame A, theCH layers are more intensely contorted and the flamesurface area is larger than in flames B and C [58]. Fur-thermore, CH LIF peak intensities differ in the threeflames. They are highest for flame B (glob = 0.75)followed by flame A (glob = 0.65) and flame C(glob = 0.55); i.e., the order is the same as for theglobal equivalence ratios. This indicates that the reac-tions occur, on average, at different local equivalenceratios and not generally around local = 1, as wouldbe assumed for diffusion flames. This observation isconfirmed by the Raman results concerning the ther-mochemical states of the flames.

3.3. Mixture fraction, temperature, and species molefractions

To yield an overview of the main features ofthe distributions of mixture fraction f , temperatureT , and mole fraction X, the results from the point-wise Raman measurements are displayed as two-dimensional charts which were obtained by interpo-lating between the measuring locations. With the op-tical setup of the Raman system, measurements wererestricted to the region h 5 mm and r 30 mm.

The distributions of the mean mixture fraction, asdisplayed in Fig. 13, show that the highest f values

are found directly above the fuel nozzle exit, as ex-pected. It is, however, remarkable that these valuesare already quite small at h = 5 mm; e.g., fmax =

n andFig. 10. Single-shot OH LIF images of flames A, B, and C.

0.072, 0.059, and 0.056 for flames A, B, and C, re-spectively. This demonstrates the fast mixing result-ing from this nozzle configuration. In comparison,the stoichiometric mixture fraction is fstoich = 0.055and the overall mixture fractions of the flames werefglob = 0.037 (A), 0.042 (B), and 0.031 (C) (see Ta-

the CH distributions (see Fig. 11 right), indicatingthat mixing and the main flame reactions are closelylinked at these heights. It is further seen that in thenear field of the nozzle, f is considerably higher inthe irz (f > fglob) than in the orz where f fglob.The relatively high f values in the irz enhance the216 P. Weigand et al. / Combustioble 1). For flames A and C, mixing is complete ath 40 mm, and for flame B, already at h 20 mm.These values are in agreement with the heights ofFlame 144 (2006) 205224stabilizing effect of the irz on the flame, because theyenable a temperature level that is higher than T glob ad.The rms values of f (displayed in Fig. 13 left) exhibit

n andFig. 11. Averaged OH LIF images (left) and CH LIF images (right).

a maximum of frms 0.04 at h = 5 mm and decreaserapidly with height. For a closer look at the mixingbehavior of the three flames, Fig. 14 shows the radialprofiles of fmean at h = 5 and h = 10 mm. Flame Areaches the largest fmean (r = 6 mm at h = 5 mm,r = 8 mm at h = 10 mm) of all three flames, which

f mean of flame B shows a significantly smaller vari-ation than that of flame C, although the mean flowfields are quite similar. The reason lies in the ther-moacoustic oscillations of flame B: In addition tothe turbulent fluctuations, the periodic variations alsocontribute to a homogenization of the time-averagedP. Weigand et al. / Combustiois explained by the high exit velocities of this flameand, thus, the shorter time for mixing before reachingh = 10 mm. It is surprising that the radial profile ofFlame 144 (2006) 205224 217mixture fraction distribution.The distributions of the mean temperatures, dis-

played in Fig. 15, reflect the different shapes of the


Fig. 12. Single-shot images of CH LIF.

flames, and it once more becomes evident that flame Bis very different from the others. It is also seen thatthe final temperatures of the flames (at large heights)are increasing with their global equivalence ratios,as expected. To identify the differences more clearly,Fig. 16 displays the axial profiles of T mean andT rms. At h = 5 mm, all three flames exhibit a similartemperature of T 1300 K. With increasing height,T mean values increase strongly, reach a maximum ath = 1020 mm, and decrease slowly afterward. Forflames A and B, the maximum mean temperatures areeven higher than T glob ad due to the relatively highf values in this region. The temperature fluctuationsreach a level of 500600 K close to the nozzle anddecrease to 4070 K at larger heights, which corre-

sponds to the inherent rms of typically 3% due tomeasurement precision. Considering the temperaturefluctuations within the irz, especially in the lower part,Flame 144 (2006) 205224

Fig. 13. Two-dimensional mixture fraction distribution (rightside: mean values, left side: rms values).

it becomes obvious that the irz is not a stationary vor-tex stable in time and space but, rather, is subject tosignificant turbulent fluctuations, as was already indi-cated by the single-shot images of OH and CH in thisregion and by the velocity fluctuations. A similar re-sult was obtained by Ji and Gore [59] in a differentswirl flame, where they showed by particle image ve-locimetry that the instantaneous structure of the irz isoften composed of a number of smaller vortices. Thismust be kept in mind for the phenomenological un-

derstanding of flame behavior.

More details of the comparison between the threeflames are seen in the radial profiles of T mean at


Fig. 14. Radial profiles of mean mixture fraction ath = 5 mm and h = 10 mm.

h = 5 and 10 mm in Fig. 17. The low-temperatureregions at r 617 mm reflect the inlet streams offresh gas. Here it is seen that flame B reaches signifi-cantly higher temperatures than the other two flames;e.g., at h = 5 mm the lowest mean temperatures areTmin = 408, 627, and 476 K for flames A, B, andC, respectively. The higher temperatures of flame Bin this region are partly explained by the more fre-quent occurrence of reactions (see CH distribution inFig. 11). However, at h = 5 mm and especially forr > 10 mm, the main source of the elevated tempera-tures is mixing of hot exhaust gas from the recircula-tion zones with fresh gas. The increased temperaturelevel at h = 5 mm enhances, of course, the reactiv-ity of the gas mixtures and, thus, the heat release andburnout [60]. This can clearly be seen in the temper-ature profiles at h = 10 mm, where flame B alreadyreaches a minimum temperature of 1007 K, whereasfor flames A and C the minimum temperatures are 554and 639 K, respectively. The transition between theinlet stream and the orz at r 20 mm is clearly vis-ible in the profile of flame A at h = 5 mm. It is alsoobvious that for h 10 mm, the temperature level inthe orz is in general lower than in the irz, which is dueto the leaner mixtures (lower f values) and heat loss

to the wall in the orz.

Phase-correlated measurements in flame B re-vealed that the phase-resolved mean temperature ofFlame 144 (2006) 205224 219

Fig. 15. Two-dimensional temperature distribution (rightside: mean values, left side: rms values).

the inflowing gas at h = 5 mm varied by about 300 Kduring an oscillation cycle [34]. This variation wascorrelated with a periodic expansion of the recircula-tion zones: When the irz penetrated into the centralair nozzle and the orz reached its maximum expan-sion, large amounts of recirculating exhaust gas weremixed into the fresh gas, increasing the temperaturewithin the inflow. The measurements further indicatedthat variations of the temperature level of the inflow

and the heat release rate were correlated, leading tothe conclusion that the temperature of the inflow hada significant influence on the heat release rate.


Fig. 16. Axial profiles of temperature (mean + rms).

Fig. 17. Radial profiles of mean temperature at h = 5 mmand h = 10 mm.

The right part of Fig. 18 displays the distributionsof the mean values of the differences between thequasi-adiabatic flame temperature T a and the mea-sured temperature T . Quasi-adiabatic flame tem-perature is defined here as the temperature for theparticular mixture fraction taken from a calculationfor a strained laminar counterflow diffusion flamewith a strain rate of a = 1 s1 [53,54]. T a has beencalculated for the measured mixture fraction at eachlocation for each single shot and from these resultsTa T has been averaged. The use of the real adi-abatic flame temperature (and composition) is not

meaningful for fuel-rich regions of turbulent flames,because the thermal decomposition of CH4 (whichis complete for adiabatic equilibrium) takes a longerFlame 144 (2006) 205224

Fig. 18. Right: two-dimensional distribution of the differ-ence between locally possible adiabatic temperature T a andthe measured mean temperature T . Left: two-dimensionaldistribution of the mean H2O mole fraction. Both values canbe taken as a measure of the reaction progress.

time than is typically available in these flames. Devi-ations between T and T a can stem either from heatloss of the flame gases, e.g., due to thermal radia-tion or wall contact, or from finite-rate chemistry ef-fects. As long as heat loss is of minor importance,the mean value of Ta T can be taken as a mea-sure of the mean reaction progress in the flame, and

is an indirect way to display the effects of finite-ratechemistry, which is discussed in more detail in theaccompanying article [32]. The results displayed in


Fig. 18 show that Ta T reaches significant valuesin all three flames; e.g., the maximum mean valuesare >1400, >900, and >1000 K for flames A, B, andC just above the nozzle exit. From the large valuesof Ta T and its distributions, it becomes obviousthat finite-rate chemistry effects play a very impor-tant role in the flames investigated. The burned gasesreach a final state with Ta T < 200 K that is closeto equilibrium. Comparison of the results reveals thatnonequilibrium effects are quite differently distrib-uted in the three flames: Flame A reaches a constantlevel of Ta T at h 55 mm, and in flame C signifi-cant effects of finite-rate chemistry are observed up toh 45 mm, whereas in flame B a uniform Ta T isattained by h 20 mm. These heights are in good ac-cordance with the CH LIF images, where for flames Band C, the same heights are found for detectable CH,and for flame A, CH is still present at the upper endof the image at h = 47 mm. The observed differencein height between flames A and C is in accordancewith the different flow velocities and Reynolds num-bers of the flames. The much faster burnout of flame Bis again related to the thermoacoustic pulsations andis probably caused mainly by the relatively high tem-perature level of the gas in the inflow as discussedbefore.

Finally, the mean distributions of the mole frac-tions X of H2O, CH4, and O2 are presented inFigs. 18 and 19. The shapes of the distributionsof X(H2O), displayed in Fig. 18 (left), resemblestrongly those of temperature for each flame (seeFig. 15). Inspection of the single-shot results (not dis-played) reveals that the correlation between X(H2O)and T is in quite good agreement with the correla-tions calculated for strained laminar flames. Fromthe single-shot correlations it is, however, seen thatthe flames experience a temperature loss in theorz, probably due to heat conduction to the burnerplate [32].

As shown in Fig. 19 (left), the distributions ofX(CH4) exhibit the highest values close to the fuelnozzle; however, for flames A and C the maximaare not exactly above the CH4 injection but shiftedslightly inward. Close to the nozzle (h < 15 mm), theCH4 distribution of flame B is significantly broaderthan those of flames A and C. A similar trend wasalready seen in the mixture fraction distributions(Figs. 12 and 13) and can also be observed for theO2 distribution (Fig. 19). This result is a further indi-cation that the periodic oscillations of the flow fieldgenerate additional mixing of fuel, air, and exhaustgas, which promotes reaction progress. The consump-tion of CH4 with increasing distance from the nozzle

is in general accordance with the decrease in Ta Texcept for a small discrepancy in the orz, whereTa T increases, due to the above-mentioned tem-Flame 144 (2006) 205224 221

Fig. 19. Left: two-dimensional distribution of the mean CH4mole fraction. Right: two-dimensional distribution of themean O2 mole fraction.

perature loss. This can be seen comparing Fig. 18(right) and Fig. 19 (left) at positions r > 25 mm andh = 010 mm. Here, the CH4 concentration is aroundor smaller than 1% but the temperature differenceTa T is larger than 200 K for flame A or even400 K for flames B and C. The temperature differ-ence is not explainable by the remaining fuel and is

probably due to heat transfer to the casing. In the irz,above h 10 mm, and in the exhaust gas region, noCH4 is found.


The distributions of X(O2) (Fig. 19, right) re-flect again the shapes of the flames and are in goodagreement with those of T and Ta T ; i.e., X(O2)decreases as T increases. The fact that the lowestmean concentrations of O2 are found near the flameaxis within the irz confirms that the mixing charac-teristics of the burner configuration generate a rel-atively fuel-rich (but still overall lean) recirculatinggas flow which is of importance for flame stabiliza-tion. The distributions of the remaining major species(N2, CO2, CO, H2) are in good agreement with theresults presented and are not displayed. The meanmole fractions of the intermediate species CO andH2 are generally below 0.018 and 0.012, respectively.The highest concentrations of these species are foundin the shear layer between the inlet flow and the irzat h = 1020 mm with near-stoichiometric mixturefractions and temperatures between 1000 and 1500 K.

4. Summary and conclusions

A laboratory-scale GT model combustor has beendescribed and three CH4/air diffusion flames with dif-ferent flame characteristics have been investigated us-ing LDV, PLIF of OH and CH, and laser Raman scat-tering. The main goals of the studies were (1) to carryout a detailed experimental analysis to improve under-standing of the physical and chemical processes lead-ing to the different behaviors of swirling flames, and(2) to provide a useful database for the verificationand improvement of numerical combustion models.The results presented in this article concern the flowfields and the flame structures, as well as the meanvalues and fluctuations of the major species concen-trations, mixture fraction, and temperature. A sec-ond article addresses turbulencechemistry interac-tions and their effect on mixing and stabilization [32].

The shapes of the mean flow fields of all threeflames are quite similar: The injected flows of CH4and air formed a cone with an opening angle /2of about 26 with respect to the flame axis. Due tothe swirl, a pronounced irz was established whichreached from h 70 mm down into the central airnozzle. This reverse flow of hot combustion productsformed the major source for the ignition and stabiliza-tion of the flames. The instantaneous flame structuresand velocity fluctuations gave evidence that the irzwas, at least in the lower part, subject to strong tur-bulent fluctuations of its shape and composition. Anorz was established in the lower part of the combus-tion chamber.

The flame structures were visualized by planar

laser-induced fluorescence of OH and CH. The re-sults showed that the instantaneous flame shapes weredominated by turbulence and that the three flamesFlame 144 (2006) 205224

could hardly be distinguished based on a single in-stantaneous OH or CH image. The reaction zonesof all three flames were generally thin (0.5 mm)and strongly wrinkled but more contorted in flame A,which had a higher Reynolds number. The averagedimages of CH PLIF showed the regions of heat re-lease. It could be seen that none of the flames wasattached to the nozzle; i.e., reactions did not startbelow h 5, 4, and 6 in flames A, B, and C, re-spectively. The flames were thus partially premixedbefore ignition. Flame B had a remarkably short flamelength of h 20 mm, whereas the other two flamesreached up to h 50 mm (flame A) and h 40 mm(flame C), which also represented a fast burnout. Theaveraged OH and CH distributions also revealed thesignificantly different shapes of the three flames. Theopening angles for the flame zones deviated from theopening angle of the flow fields for flame C and espe-cially for flame B, while for flame A the two anglesmatched roughly.

While a certain similarity was seen for flames Aand C, with opening angles /2 of 30 and 45, re-spectively, flame B had a significantly larger openingangle (/2 = 75) and was much shorter. The differ-ent shapes of flames B and C were surprising becausetheir mean flow fields were very similar, especiallyat the nozzle exit. Except for the velocity fields, allother measured quantities confirmed that flame B ex-hibited a different behavior than flames A and C. Mi-crophone measurements revealed that the sound emis-sions of flame B were concentrated at a frequency of290 Hz, proving the occurrence of thermoacousticoscillations, and previously reported phase-resolvedLDV and PLIF measurements showed significant pe-riodic variations of the flame structure during an os-cillation cycle. Periodic changes in the expansion ofthe inner and outer recirculation zones enhanced themixing of fuel, air, and exhaust gas, which in turn con-tributed to increased reaction progress.

Measurements of the mixture fraction demon-strated the fast mixing of fuel and air of this nozzleconfiguration. At h = 5 mm, variations of the meanmixture fraction along the radial profile remained be-tween fmean = 0.02 and 0.08 for flame A and wereeven smaller for the other flames. Mixing was com-pleted at about the same height as the heat release, i.e.,at h 40 mm for flames A and C and at h 20 mmfor flame B. Within the irz, f and T were higher andX(O2) was smaller than the global values. The flowand the mixing characteristics of the burner enhancedthe effect of the irz with respect to ignition and stabi-lization of the flame. Although measurements could

not be performed below h = 0 mm, the results ob-tained indicated that hot combustion products weretransported via the irz into the central nozzle, where


they mixed with air before reentering the combustionchamber.

The distributions of Ta T showed that in allthree flames, pronounced finite-rate chemistry effectsoccurred which led to a significant deviation fromequilibrium composition and temperature, especiallywithin the high-velocity regions of the inlet flow andthe neighboring shear layers. Finally, it should benoted that the experimental data are available fromone of the authors (W.M.) on request for validation ofnumerical codes for swirling turbulent flames.

Acknowledgments

The work presented here was performed mainly inthe frame of the project Combustion Control and Sim-ulation, funded by the State of Baden-Wrttemberg,and as part of the DLR project NACOS. The finan-cial support within these projects is gratefully ac-knowledged by the authors. We furthermore thankB. Lehmann for execution of the LDV measurementsand B. Noll for fruitful discussions.

References

[1] A.K. Gupta, D.G. Lilley, N. Syred, Swirl Flows, Aba-cus Press, Kent, 1984.

[2] N. Syred, N.A. Chigier, J.M. Ber, Proc. Combust.Inst. 13 (1971) 617624.

[3] N. Syred, J.M. Ber, Combust. Flame 23 (1974) 143201.

[4] R. Weber, J. Dugu, Prog. Energy Combust. Sci. 18(1992) 349367.

[5] S.M. Correa, Proc. Combust. Inst. 27 (1998) 17931807.

[6] A.H. Lefebvre, Gas Turbine Combustion, Taylor &Francis, Philadelphia, 1999.

[7] H.J. Bauer, Prog. Comput. Fluid Dyn. 4 (2004) 130142.

[8] W. Meier, X.R. Duan, P. Weigand, B. Lehmann,Gaswrme Int. 53 (2004) 153158.

[9] W. Stricker, in: K. Kohse-Hinghaus, J. Jeffries (Eds.),Applied Combustion Diagnostics, Taylor & Francis,New York, 2002, pp. 155193.

[10] A.C. Eckbreth, Laser Diagnostic for Combustion Tem-perature and Species, Gordon & Breach, 1996.

[11] K. Kohse-Hinghaus, J. Jeffries (Eds.), Applied Com-bustion Diagnostics, Taylor & Francis, New York,2002.

[12] A.R. Masri, R.W. Dibble, R.S. Barlow, Prog. EnergyCombust. Sci. 22 (1996) 307362.

[13] J. Wolfrum, Proc. Combust. Inst. 28 (1998) 141.[14] H. Kaaling, R. Ryden, Y. Bouchie, D. Ansart, P. Ma-gre, C. Guin, in: 13th International Symposium on AirBreathing Engines (ISABE), Chattanooga, TN (USA),1997.Flame 144 (2006) 205224 223

[15] S. Kampmann, T. Seeger, A. Leipertz, Appl. Opt. 34(1995) 27802786.

[16] F. Dinkelacker, A. Soika, D. Most, D. Hofmann, A.Leipertz, W. Polifke, K. Dbbeling, Proc. Combust.Inst. 27 (1998) 857865.

[17] R. Fink, A. Hupfer, D. Rist, in: Proceedings, ASMETurbo Expo 2002, GT-2002-30078.

[18] C.S. Cooper, N.M. Laurendeau, Appl. Phys. B 70(2000) 903910.

[19] C.S. Cooper, N.M. Laurendeau, Combust. Flame 123(2000) 175188.

[20] W.-P. Shih, J.G. Lee, D.A. Santavicca, Proc. Combust.Inst. 26 (1996) 27712778.

[21] Y. Deguchi, M. Noda, Y. Fukuda, Y. Ichinose, Y. Endo,M. Inida, Y. Abe, S. Iwasaki, Meas. Sci. Technol. 13(2002) R103R115.

[22] P.O. Hedman, D.L. Warren, Combust. Flame 100(1995) 185192.

[23] S.-Y. Lee, S. Seo, J.C. Broda, S. Pal, R.J. Santoro, Proc.Combust. Inst. 28 (2000) 775782.

[24] A. Arnold, R. Bombach, W. Hubschmid, B. Kppeli,ERCOFTAC Bull. 38 (1998) 1019.

[25] J. Fritz, M. Krner, Th. Sattelmayer, in: Proceedings,ASME Turbo Expo 2001, 2001-GT-0054.

[26] C. Lfstrm, J. Engstrm, M. Richter, C.F. Kaminsky,P. Johansson, K. Nyholm, J. Nygren, M. Aldn, in: Pro-ceedings, ASME Turbo Expo 2000, 2000-GT-0124.

[27] C.M. Gittins, S.U. Shenoy, H.R. Aldag, D.P. Pacheco,M.F. Miller, M.G. Allen, in: 38th AIAA Aerospace Sci-ences Meeting, Reno, NV, 2000.

[28] U.E. Meier, D. Wolff-Gamann, J. Heinze, M. Froder-mann, I. Magnusson, G. Josefsson, in: 18th Interna-tional Congress on Instrumentation in Aerospace Simu-lation Facilities (ICIASF 99), Toulouse, 1999, pp. 7.17.7.

[29] U.E. Meier, D. Wolff-Gamann, W. Stricker, Aero-space Sci. Technol. 4 (2000) 403414.

[30] M. Carl, T. Behrendt, C. Fleing, M. Frodermann, J.Heinze, C. Hassa, U. Meier, D. Wolff-Gamann, S.Hohmann, N. Zarzalis, ASME J. Eng. Gas TurbinesPower 123 (2001) 810816.

[31] O. Kunz, B. Noll, R. Lckerath, M. Aigner, S.Hohmann, in: 37th AIAA/ASME/SAE/ASEE JointPropulsion Conference and Exhibition, Salt Lake City,UT, 2001, AIAA 2001-3706.

[32] W. Meier, X. Duan, P. Weigand, Combust. Flame 144(2006) 225236.

[33] R. Giezendanner, O. Keck, P. Weigand, W. Meier, U.Meier, W. Stricker, M. Aigner, Combust. Sci. Tech-nol. 175 (2003) 721741.

[34] X.R. Duan, W. Meier, P. Weigand, B. Lehmann, Appl.Phys. B 80 (2005) 389396.

[35] P. Weigand, W. Meier, X.R. Duan, R. Giezendanner-Thoben, U. Meier, Flow Turbulence Combust., in press.

[36] M. Cao, H. Eickhoff, F. Joos, B. Simon, in: ASMEPropulsion and Energetics, 70th Symposium, AGARD

Conf. Proc. 422 (1987) 8.1.

[37] W. Leuckel, N. Fricker, J. Inst. Fuel 49 (1976) 103112.[38] D.G. Lilley, AIAA J. 15 (8) (1977) 10631078.

224 P. Weigand et al. / Combustion and Flame 144 (2006) 205224

[39] J.A. Sutton, J.F. Driscoll, Appl. Opt. 42 (2003) 28192828.

[40] J. Tobai, T. Dreier, J.W. Daily, J. Chem. Phys. 116(2002) 40304038.

[41] G.H. Dieke, H.M. Crosswhite, J. Quant. Spectrosc. Ra-diat. Transfer 2 (1962) 97199.

[42] J. Luque, D.R. Crosley, J. Chem. Phys. 104 (1996)39073913.

[43] J. Luque, D.R. Crosley, SRI Report No. MP 98-021,SRI Int., Menlo Park, CA, 1998.

[44] C.D. Carter, J.M. Donbar, J.F. Driscoll, Appl. Phys.B 66 (1998) 129132.

[45] J. Luque, R.J.H. Klein-Douwel, J.B. Jeffries, D.R.Crosley, Appl. Phys. B 71 (2000) 8594.

[46] O. Keck, W. Meier, W. Stricker, M. Aigner, Combust.Sci. Technol. 174 (2002) 73107.

[47] P. Weigand, R. Lckerath, W. Meier, http://www.dlr.de/VT/Datenarchiv, 2003.

[48] V. Bergmann, W. Meier, D. Wolff, W. Stricker, Appl.Phys. B 66 (1998) 489502.

[49] G. Dixon-Lewis, Proc. Combust. Inst. 23 (1990) 305324.

[50] R.S. Barlow, R.W. Dibble, J.-Y. Chen, R.P. Lucht,Combust. Flame 82 (1990) 235251.

[51] J. Luque, R.J.H. Klein-Douwel, J.B. Jeffries, G.P.Smith, D.R. Crosley, Appl. Phys. B 75 (2002) 779790.

[52] H.N. Najm, P.H. Paul, C.J. Mueller, P.S. Wyckoff,Combust. Flame 113 (1998) 312332.

[53] J.-Y. Chen, University of California at Berkeley, privatecommunication.

[54] J.H. Miller, R.J. Kee, M.D. Smooke, J.F. Grcar, PaperWSS/CI 84-10, in: Western State Section of the Com-bustion Institute, Spring Meeting, 1984.

[55] U. Rahmann, A. Btler, U. Lenhard, R. Dsing,D. Markus, A. Brockhinke, K. Kohse-Hinghaus,LASKINA Simulation Program for Time-ResolvedLIF-Spectra, International Report, University of Biele-feld, Faculty of Chemistry, Physical Chemistry I,http://pc1.uni-bielefeld.de/laskin V. 5.0.

[56] K.J. Rensberger, M.J. Dyer, R.A. Copeland, Appl.Opt. 27 (1988) 36793689.

[57] R. Giezendanner-Thoben, U. Meier, W. Meier, M.Aigner, Flow Turbulence Combust., in press.

[58] W. Meier, X.R. Duan, P. Weigand, Proc. Combust.Inst. 30 (2005) 835842.

[59] J. Ji, J.P. Gore, Proc. Combust. Inst. 29 (2002) 861867.[60] M. Braun-Unkhoff, DLR Stuttgart, private communica-

tion.

Investigations of swirl flames in a gas turbine model combustorIntroductionExperimentalCombustor and flamesMeasuring techniques

Results and discussionLDV measurementsFlame structures from OH LIF and CH LIF measurementsMixture fraction, temperature, and species mole fractions

Summary and conclusionsAcknowledgmentsReferences

Documents

CF Weigand Et Al