Available online at www.sciencedirect.comProceedings

Proceedings of the Combustion Institute 33 (2011) 2937–2944

www.elsevier.com/locate/proci

of the

CombustionInstitute

Turbulent flame speed for syngas at gas turbinerelevant conditions

S. Daniele a,*, P. Jansohn a, J. Mantzaras a, K. Boulouchos b

a Paul Scherrer Institut (PSI), Combustion Research Laboratory, 5232 Villigen PSI, Switzerlandb Swiss Federal Institute of Technology (ETH), Aerothermochemistry and Combustion Systems Laboratory

Sonneggstr. 3, 8092 Zurich, Switzerland

Available online 7 August 2010

Abstract

Modifications of conventional natural-gas-fired burners for operation with syngas fuels using lean pre-mixed combustion is challenging due to the different physicochemical properties of the two fuels. A keydifferentiating parameter is the turbulent flame velocity, ST, commonly expressed as its ratio to the laminarflame speed, SL. This paper reports an experimental investigation of premixed syngas combustion at gasturbine like conditions, with emphasis on the determination of ST/SL derived as global fuel consumptionper unit time. Experiments at pressures up to 2.0 MPa, inlet temperatures and velocities up to 773 K and150 m/s, respectively, and turbulence intensity to laminar flame speed ratios, u0/SL, exceeding 100 are pre-sented for the first time. Comparisons between different syngas mixtures and methane clearly show muchhigher ST/SL for the former fuel. It is shown that ST/SL is strongly dependent on preferential diffusive-ther-mal (PDT) effects, co-acting with hydrodynamic effects, even for very high u0/SL. ST/SL increases with ris-ing hydrogen content in the fuel mixture and with increasing pressure. A correlation for ST/SL valid for allinvestigated fuel mixtures, including methane, is proposed in terms of turbulence properties (turbulenceintensity and integral length scale), combustion properties (laminar flame speed and laminar flame thick-ness) and operating conditions (pressure and inlet temperature). The correlation captures effects of prefer-ential diffusive-thermal and hydrodynamic instabilities.� 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Syngas combustion; Turbulent flame speed; High pressure; High inlet temperature; Lean premixed flame

1. Introduction

Lean premixed combustion represents thestate-of-the-art gas turbine combustion technol-ogy for high efficiency and low emissions. Thistechnology has been optimized for natural gascombustion and has demonstrated single-digit

1540-7489/$ - see front matter � 2010 The Combustion Institdoi:10.1016/j.proci.2010.05.057

* Corresponding author. Fax: +41 563102199.E-mail address: salvatore.daniele@psi.ch (S. Daniele).

nitrogen oxides (NOX) and very low carbon mon-oxide (CO) emissions (<20 ppm).

In recent times, integrated gasification com-bined cycle (IGCC) power plants have attractedincreased attention, predominantly due to two rea-sons. The main driving force is the availability ofdiverse fuels, like biomass, coal, tars, etc. (sourcesof syngas), and the increasing concern for fuel sup-ply security and flexibility, which prompt the useof syngas for high efficiency gas turbine-basedpower generation. A second reason is that syn-gas-based power plants facilitate cost-effective

ute. Published by Elsevier Inc. All rights reserved.

2938 S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944

carbon dioxide (CO2) emissions reduction, whencombined with fuel decarbonization.

The gasification of the aforementioned diversefuels leads to a mixture, termed syngas, whosecomposition varies according to the gasificationprocess and the original feedstock. Main fuelcomponents in syngas are hydrogen (H2) andCO, while diluents such as water (H2O), nitrogen(N2) or CO2 can also be present in variousamounts. These low-to-medium heating valuegases are characterized by chemical and physicalproperties that can be drastically different fromthe ones of natural gas. High values of laminarflame speed (SL), low density, high mass diffusiv-ity and consequently low Lewis numbers (thermalover mass diffusivity), require different engineoperating conditions for lean syngas combustion(so as to achieve a similar flame temperature withmethane). The resulting effects on the turbulence-chemistry interactions portray a quite differentcombustion system when employing syngas fuels.

One of the most important parameters in pre-mixed turbulent combustion is the turbulent flamespeed (ST), with its knowledge being of para-mount importance for combustor design. The sig-nificance of mixture composition has beenexemplified by Lipatnikov and Chomiak in theirreview [1], concluding that the turbulent flamespeed is strongly affected by molecular transport:in fuel-lean syngas/air mixtures, transport effectsare dominated by the H2 component.

As reviewed in [2], in contrast to the extensiveknowledge gained in the lasts decades for hydro-carbons (especially CH4), very few experimentaldata are available in the literature for syngas fuels[3–5]. Moreover, the aforementioned data are notdirectly comparable due to the different adopteddefinitions for ST, and the different experimentalmethodologies used. The lack of syngas databecomes very obvious especially when consideringgas turbine relevant conditions, involving highpressures (1–3 MPa), high preheat temperatures(600–700 K) and high turbulence intensities (u0/SL > 50).

The objective of this work is to describe turbu-lent combustion characteristics and in particularturbulent flame speeds for a variety of syngas-based fuel mixtures as function of equivalenceratio, pressure, inlet temperature and velocity,and turbulence intensity at gas turbine pertinentconditions.

2. Experimental setup, measuring techniques anddata processing

Experiments are performed in a high pressure,optically accessible cylindrical combustion cham-ber (see Fig. 1) delivering a maximum thermalpower of 400 kW. The chamber has a length of320 mm with an inside diameter of 75 mm, and

it is made of an air-cooled double-wall quartz tube(for more details, see [6]). The fuel/air mixture isdelivered via a 25 mm diameter tube, coaxial tothe reactor. The fuel is injected and mixed withelectrically-preheated air 400 mm upstream ofthe combustion chamber. Fuel injection is accom-plished in a co-flow configuration with the pre-heated air, through an array of five jets (1 mmin diameter) uniformly distributed within thedelivery tube.

A sudden radial expansion by a factor of threebetween the delivery tube and the burner inducesan outer recirculation zone which stabilizes theflame. There is no swirl or piloting flame applied.Pressurization is achieved by a cylindrical tankenclosing the reactor and delivery tube. Opticalaccess to the reaction zone is provided by threequartz windows positioned on the tank, two sideones and one along the reactor axis (see Fig. 1).

Planar laser induced fluorescence (PLIF) of theOH radical was employed to assess the flame frontposition and turbulent flame speeds. The detailedlaser/camera setup is described in [7] (excitationbeam at 285 nm through the use of a pulsedNd:YAG/dye laser system with 30 ns pulse dura-tion, and fluorescence collection at 308 and314 nm). The 285 nm PLIF light sheet enteredthe reactor counterflow, from the rear tank axialwindow. The raw images were post-processedadopting the methodology proposed in [8] withthe modifications described in [5]. These modifica-tions, introducing a local threshold methodologyfor the flame front location, were necessary toprocess PLIF images at high pressures and ultra-lean stoichiometries.

Measurements are reported in terms of param-eters such as turbulent flame speed, ST, flamelength, FL, and flame brush thickness, fBT. Abso-lute values of ST were derived by solving the con-tinuity equation at the inlet section and at theaverage flame front surface [5]. The flame surfacehas been defined at a reaction progress variable ofhci = 0.05, with c denoting OH signal intensity.Therefore, ST represents a “volumetric, global fuelconsumption rate” [2,9,10]. The flame lengthF hciL ¼ 0:05 is defined as the axial distance betweenthe combustor inlet and the position of the aver-age flame front (hci = 0.05) along the centerline.Twice the difference of flame front positions athci = 0.50 and hci = 0.05 defines the flame brushthickness, fBT = 2(F hci¼0:50

L � F hci¼0:05L ).

For each operating condition, the final results(ST, F hciL ¼ 0:05, fBT) are based on 400 images(instantaneous shots) acquired at a 20 Hz sam-pling rate. An ST variation of up to ±0.5 m/shas been observed for repeated measurements atany experimental condition. In general, it can bestated that high pressure/ultra-lean flames arecharacterized by a larger ST uncertainty whencompared to low pressure/less-lean flames. Thisis because high pressure and/or very lean mixtures

Fig. 1. Schematic of test rig. Fig. 2. Borghi diagram (modified by Peters) with datapoints acquired in this work. Karlowitz (Ka, Kad),Damkholer (Da) and turbulent Reynolds (ReT) numbersare based on integral length scale (LT). Legend as inTable 2.

Table 1Composition of fuel stream (oxidizer stream is air).

S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944 2939

lead to a decrease in the PLIF intensity that inturn affects the quality of the results during theflame front detection procedure [5].

Flow field characteristics including turbulenceintensity, u0 and integral length scales, LT havebeen measured in past works by particle imagevelocimetry, PIV [11].

Unstrained laminar flame speeds, SL, werecomputed using the GRI 3.0 mechanism [12],while characterization of strained SL will bereported in a future work. Table 1 provides thecomposition coding for Figs. 2, 4, and 7 andTable 2 the pressure coding for Figs. 2, 4, 6, and 7.

3. Results and discussion

3.1. Turbulent and chemical kinetic regimes of dataset

Figure 2, positions the data of the presentwork in the Borghi diagram, as modified by Peters[13]. The bulk of the data fall in the thin reactionzones regime, a major part characterized byDamkholer numbers (Da) <1, with only a fewdata points in the broken reaction zones, charac-terized by Karlovitz number (Kad) >1. For CH4

no data could be acquired at Da < 1 as theseflames encountered lean blow-out in this regime.In Fig. 2, lm denotes the Zimont mixing scale [14].

Figure 3 provides OH-PLIF single-shot imagesat different equivalence ratios, U. An elongation

of the flame length is observed with decreasingU, while no differences in the flame morphologyare evident when crossing Da = 1. The probabilityof a broken flame surface and the presence ofburning isolated pockets increase with decreasingU. For Kad > 1 the reaction zone is broken; how-ever, analysis of CO emissions at the exhaust con-firms that combustion is still complete. Flameimages for Kad > 1 have been processed with thesame methodology, as it was always possible todefine an average continuous reaction zone.

3.2. Relation between turbulent flame speed andflame front position

Recently [2,10,15], it has been stressed that tur-bulent flame speeds may depend on the geometryof the particular experimental rig, since localflame front wrinkles bear the memory of the onesgenerated upstream. In this respect, flames gener-ated with similar geometries should preferably becompared. To eliminate most of the geometrydependence, we correlate turbulent flame speedvalues, ST, with turbulence characteristics suchas integral length scale, LT, and turbulence inten-sity, u0, measured under atmospheric-pressure,non-reactive flow conditions at the flame location.

The turbulent flow field in our configurationhas been studied in the past [11]. Several experi-ments were conducted with different perforatedplates mounted in the fuel/air mixing section. Inthe current study no perforated plates were used:it was impossible to position any element in themixing section (which could act as potential flameholder) when firing syngas because of its highflashback propensity as described in more detailsin [16]. Previous studies have shown that the effectof different types of perforated plates on the tur-bulent flow field is only restricted to core flowregions close to the combustor inlet (x < 2d,

Fig. 3. OH-PLIF images corresponding to three differ-ent regimes.

Fig. 4. (a) ST, u0, LT, fBT versus F hci¼0:05L /d, and (b)

averaged OH-PLIF images for two flames. Legend as inTable 2.

Fig. 6. ST/SL versus u0/SL: pressure effect for syngas(50% vol. H2/50% vol. CO). Legend as in Table 2. Thedashed-dotted lines refer to constant SL values.

Fig. 5. ST/SL versus u0/SL: H2 content effect for syngasand pressure effect for CH4.

2940 S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944

d = diameter of fuel/air delivery pipe), revealingthat the turbulence field at the location of theflame front is essentially dominated by theshear-generated turbulence due to the backwardfacing step geometry [11]. Therefore, turbulenceintensity and integral length scales were inferredin the present study by averaging over four differ-ent flow conditions corresponding to four perfo-rated plate geometries. The resulting averagevalues of u0 and LT on the combustor symmetryaxis are presented in Fig. 4 versus the flame frontposition, F hci¼0:05

L normalized by the injector diam-

eter, d. The variation of the centerline u0 in each ofthe four cases with respect to the average value,was at most 1 m/s. Centerline values are consid-ered representative of the turbulence characteris-tics along the entire flame front: it has beenshown that the flames stabilize along contours ofnearly constant u0 and LT [11].

Figure 4a presents a wide selection of ST mea-surements comprising inlet temperatures from 423to 773 K, equivalence ratios between 0.25 and0.75, pressures from 0.1 to 2.0 MPa, and variousfuel mixtures listed in Table 1. The strategy forthe selection of these fuel compositions is detailedin [5]. In the same figure, fBT values are alsoplotted.

The dashed and the dashed–dotted-lines inFig. 4a represent ST values corresponding to aflame with an ideal conical shape and to a rota-tionally-symmetric ellipsoid, respectively. The

correlation between ST and F hci¼0:05L is largely inde-

pendent of gas mixture composition and operat-ing conditions. “Short” flames assume conical

Fig. 7. Normalized turbulent flame speed data. Legend as in Tables 1 and 2.

Table 2Legend for pressures.

Symbol P (MPa) Symbol P (MPa)

j 0.10 � 1.25N 0.25 d 1.50. 0.50 h 1.75I 0.75 4 2.00J 1.00

S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944 2941

shapes, while “longer” flames rather resemble axi-symmetric ellipsoidal shapes as depicted in Fig. 4bby the average OH-PLIF images (the arrows con-nect the data points to the corresponding flame

locations, F hci¼0:05L ). Moreover, Fig. 4a, shows that

the angle method used by Kobayashi [17] andother authors [4] for the determination of ST,can lead for our combustor geometry to errorsnot only in the absolute value of ST but also inthe observed trends (see non-constant differencebetween the dashed and the solid lines). Venk-ateswaran et al. [4], using the same configurationas Kobayashi, reported an error by the anglemethod in estimating absolute values of ST, stat-ing however that trends were not affected. Indirect contrast, Kobayashi showed no differencebetween values calculated with the angle methodand the progress variable method (hci = 0.1)[18]. With the approach presented in this paper,it is possible to generate data at different u0 (atthe same inlet velocity) without varying the over-all combustor geometry. The flame stabilizes at acertain location, where it encounters specific val-ues of u0 and LT, according to its propagationcapability.

Until now there are no values of global fuelconsumption rate, ST, reported in the literaturefor syngas compositions at high preheating tem-peratures, high pressures and high turbulenceintensities. In [4], ST (global consumption) valuesfor only atmospheric-pressure and temperatureare reported. A one-to-one comparison with thiswork is not possible as the conditions are different

and the methodology used to post-process thedata is also different.

3.3. Mixture composition effect on turbulent flamespeed

Figure 5 reports measurements for a constantpressure of 0.5 MPa, To = 623 K and U0 = 40 m/sfor four different fuel mixtures. An additional data-set for CH4 at 0.25 MPa is provided. Data pointscharacterized by the same U are connected by addi-tional solid lines.

It is well known that ST/SL values exhibit a firstzone, for low values u0/SL, wherein ST/SL is almostlinearly proportional to u0/SL, and a second zonewhere the ST/SL versus u0/SL curve bends andappears to level-off as reviewed in [9]. The bendingoccurs for a particular u0/SL depending on fuelmixture composition. Data for CH4 have beenapparently acquired within this second zone only.

The effect of pressure on ST for CH4 is depictedby the difference between the two data sets indi-cated with open and filled triangles. Increase inST/SL with rising pressure has already beenreported in [17,18]. This effect was attributed[17] to the growth of the hydrodynamic instability,DL, after Darrieus and Landau [19] whodescribed this instability for laminar flamelets.

It is noted that Kobayashi’s findings [17] wererecently questioned by Lipatnikov and Chomiak[20], arguing that there is no evidence for theDL instability being alone responsible for theincrease in ST/SL with rising pressure. Moreover,they concluded [1] that DL instability can be sig-nificant only for weak turbulence. The decreaseof importance of DL effects at high u0/SL is furthersupported by Peters [21], although he provided noinsights on the impact of pressure. Therefore, theeffect of DL instabilities at elevated pressures isstill an open question.

Data at constant pressure confirm trends pre-sented elsewhere at significantly lower u0/SL andreviewed in [1]: the higher the H2 content in the

2942 S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944

mixture, the larger the ST/SL. This behavior canbe attributed to preferential diffusive-thermalinstability, PDT, as described in [1].

As u0 in this work is taken at the flame frontposition along the centerline, and since u0 variesupon U, it is not possible to acquire data at con-stant U and u0/SL simultaneously. Therefore, weverified numerically that the Lewis number,LeH2

, for every fuel mixture, does not changeappreciably within the range of equivalence ratiospresented, such that the difference in ST/SL forP = 0.50 MPa (Fig. 5) can be predominantlyattributed to PDT effects due to different absoluteH2 levels in the mixture.

3.4. Pressure effect on turbulent flame speed

Figure 6 depicts data at different pressures(0.1–2.0 MPa) and various U (0.30–0.50), for a50% vol. H2/50% vol. CO composition. In addi-tion, linear regression lines for constant U(dashed-lines) and constant SL (dashed–dotted-lines) are shown. Regression lines for constant u0

roughly follow the U = constant lines, and arethus not shown. This behavior is based on theexperimental observation that the flame frontposition for constant U is only weakly affectedby pressure, thus the reaction zone encountersroughly the same u0 (see Fig. 4a).

An increase in pressure leads to an increase inST/SL, irrespective of the way the pressureincrease is attained (at constant u0/SL, constantu0, constant SL, or constant U). This trend is sim-ilar to that reported for CH4 [17] and alsoobserved in the present data (Fig. 5). As a conse-quence, the phenomenological explanation forCH4 [17], invoking the role of DL instability forthe increase of ST/SL with pressure, could be con-sidered valid for syngas as well, however, at higheru0/SL. The increase in ST/SL with rising pressure isstronger when moving along a U isoline. In thiscase u0 is also kept roughly constant, such thatthe increase in u0/SL can be studied solely due tothe decrease in SL. Due to the decrease in SL,for constant u0 the flame front becomes more sen-sitive to hydrodynamic wrinkling, leading to theaccelerated increase in ST/SL.

In case of syngas (LeH2< 1), we expect PDT

effects to be important and to further discriminatedata acquired at different pressures. To evaluatethe role of PDT effects, we compare the ST/SL

ratios at two pressures, 0.25 and 0.50 MPa forthe syngas mixture presented in Fig. 6 and forthe CH4 in Fig. 5. Comparisons are performedwithin a range of u0/SL (20–40) whereby ST/SL isweakly dependent on u0/SL for syngas, and fairlyindependent for CH4. These ratios are roughly2.5 for syngas and 1.6 for CH4. Hence, it appearsthat the syngas mixture undergoes an additionalenhancement of ST/SL, most probably attribut-able to PDT and to the coupling between DL

and PDT instabilities. The difference in thesetwo factors cannot be attributed to the differencein laminar flame thickness; in fact, within therange of u0/SL, considered for this comparison,the syngas mixture is characterized by a largerthickness, due to the smaller U (larger flame thick-ness reduces the possibility of flame wrinkling atsmall scales and does not contribute to an increasein ST/SL). An account for different flame thick-nesses is addressed in the next section. A detaileddescription of the pressure effect on SL alone hasbeen reported in [16].

3.5. Correlation for normalized turbulent flamespeed

Figure 7 reports a database of over 300 mea-sured operating conditions (the data notation isgiven in Tables 1 and 2).

On the abscissa, the terms P/Po and To/Taccount for pressure and inlet temperature varia-tion (we chose: Po = 0.1 MPa, To = 1 K, as in[22]). The term LT/dL accounts for the capabilityof eddies having size between dL and LT to wrin-kle the flame front at a given u0; this term appearsto also capture Lewis number effects via thedependence of dL on the thermal diffusivity.

A first data set, where ST/SL increases with awell-defined power law (proportional range, seeinsert), and a second one where ST/SL valueslevel-off (bending range) can be distinguished.The bending occurs at u0/SL values dependenton pressure, and not on mixture composition.The onset of bending is found to depend on theratio of the following time scales:

sf ¼LT

u0and sc ¼

fBT

STð1Þ

where sf is the turnover time of an eddy having thesize of the integral length scale, thus representingthe longest turbulent timescale interacting withthe flame front; sc is the time necessary for theflame front to travel through the flame brushthickness, fBT, thus denoting the timescale atwhich the flame can respond to flow perturbations.

Eddies characterized by sf much shorter thansc, have too high frequencies (1/sf) to interactand contort the flame front, thus their contribu-tion to the wrinkling process and consequentlyto the increase of ST/SL is meager or negligible.This concept was demonstrated in [23,24]. The cri-terion sc = 0.1sf appears to be appropriate to sep-arate the dataset in a “linear log–log” behavior(see insert in Fig. 7 and a “bending” one. As amatter of fact, this criterion (sc = 0.1sf) roughlycorresponds to ST = u0 (as it was found thatfBT � 10LT).

The dashed-line in Fig. 7 describes the trend ofthose data in the “linear” range with comparablesf and sc (sc/sf � 0.4). Above this limiting line it

S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944 2943

is not possible to find any data point: at these con-ditions, eddies can wrinkle the flame front appar-ently with maximum efficiency and no furtherwrinkling (i.e. acceleration of fuel consumption)is possible. A correlation for the dashed-line datain Fig. 7 can be written as:

ST

SL¼ a

u0

SL

� �0:63 LT

dL

� ��0:37 PP o

� �0:63 TT o

� ��0:63

;

a ¼ 337:5 ð2ÞThe comparison of this correlation with otherones available in the literature leads to the follow-ing considerations:

1. A dependence of ST/SL on u0/SL with a similarexponent (n = 0.7) was also reported by Kli-mov [25].

2. The negative exponent n = �0.37 of LT/dL isnot in contrast with other correlations[13,26]; for the data set supporting the abovecorrelation, this term is related to P/Po as:

LT

dL

� �¼ b

PP o

� �0:66

; b ¼ 8:3 ð3Þ

By substituting Eq. (3) in (2), the exponent of LT/dL appears to be positive (0.58), in agreement withthe previous mentioned correlations [13,26].3. Kobayashi [18,27] has proposed similar corre-

lations, accounting just for the pressure ratio.We also attempted simpler correlations by plot-ting ST/SL versus (u0/SL)(P/Po)(To/T), notreported here for brevity. This correlation pro-duced also good results; however data pointscharacterized by large differences in LT/dL werenot correlating on a single straight line as inFig. 7 but on parallel lines having roughly thesame exponent, with proportionality coeffi-cients being higher for lower pressures. Thesedifferences were particularly obvious for dataat 0.1 MPa (as depicted by Fig. 2), which arecharacterized by much smaller LT/dL. In[18,27], Kobayashi obtained a correlation forST/SL versus (u0/SL)(P/Po) with pressure powerlaws having exponents n = 0.38 and 0.40. Thesemeasurements were performed within the cor-rugated flamelets regime. Working out the glo-bal exponent, n, in P/Po by substituting againEq. (3) in (2), we obtain n = 0.385. The agree-ment with [18,27] is seen as a confirmation thatthere is no experimental evidence that the tworegimes (corrugated flamelets and thin reactionzones) differ when comparing the ST, asobserved in [1].

4. Conclusions

Lean premixed turbulent flames were investi-gated at gas turbine relevant conditions. Extensivemeasurements were performed for a variety of

syngas mixtures and for methane. Measurementswere mainly conducted in the thin reaction zonesregime. For the first time, experimental turbulentflame speed data were reported for syngas at pres-sures up to 2.0 MPa, preheat temperatures up to773 K and u0/SL > 100. In contrast to many otherworks, ST in terms of global consumption has beenrelated to the turbulence properties measured atthe flame front and not at the inlet section, so asto minimize the dependence of this parameter onthe geometrical characteristics of the burner.

The dependence of ST/SL on H2 content in thefuel mixture and on pressure was assessed. Theincrease of ST/SL on H2 content was attributedto preferential diffusive-thermal (PDT) effects,being active also at high turbulence intensity.These effects have been seen to also play a rolein the increase of ST/SL with pressure, attributedto the coupling of PDT with hydrodynamicinstabilities.

Syngas ST/SL versus u0/SL data show linearand bending characteristics, similar to that knownfor CH4. The onset of bending depends on theratio between the integral length scale eddy turn-over time and the timescale at which the flamefront can react to flow perturbations. The latterwas expressed as the ratio of the flame brushthickness to ST. A correlation for ST/SL indepen-dent on the fuel mixture composition, (appar-ently) capable of capturing both hydrodynamicand preferential diffusive-thermal instabilities hasbeen proposed.

Acknowledgments

We gratefully acknowledge financial supportof this research by the Swiss Federal Office of En-ergy (BFE) and Alstom Power Switzerland, andthank Dr. Rolf Bombach, Dr. Wolfgang Kreut-ner, Dr. Alexey Denisov and Mr. Daniel Ernefor supporting the experimental campaign.

References

[1] A. Lipatnikov, J. Chomiak, Prog. Energy Combust.Sci. 31 (2005) 1–73.

[2] R.K. Cheng, Synthesis Gas Combustion – Funda-mentals and Applications, CRC Press, Roca Raton,London, New York, 2009, pp. 129–168.

[3] R.K. Cheng, D. Littlejohn, P.A. Strakey, T.Sidwell, Proc. Combust. Inst. 32 (2009) 3001–3009.

[4] P. Venkateswaran, D. Marshall, D.R. Noble, J.Antezana, J.M. Seitzman, T.C. Lieuwen, Globalturbulent consumption speeds (ST,GC) of H2/COblends, Proc. ASME Turbo Expo 2009, Orlando,Florida, 2009.

[5] S. Daniele, P. Jansohn, K. Boulouchos, Flame frontcharacteristic and turbulent flame speed of leanpremixed syngas combustion at gas turbine relevantconditions, Proc. ASME Turbo Expo 2009, Orlan-do, Florida, 2009.

2944 S. Daniele et al. / Proceedings of the Combustion Institute 33 (2011) 2937–2944

[6] P. Griebel, E. Boschek, P. Jansohn, J. Eng. GasTurb. Power-Trans. ASME 129 (2007) 404–410.

[7] P. Griebel, R. Bombach, A. Inauen, W. Kreutner,R. Scharen, Structure and NO emission of turbulenthigh pressure lean premixed methane/air flames,Sixth European Conference on Industrial Furnacesand Boilers, INFUB, 2002, pp. 45–54.

[8] P. Griebel, P. Siewert, P. Jansohn, Proc. Combust.Inst. 31 (2007) 3083–3090.

[9] A. Lipatnikov, J. Chomiak, Prog. Energy Combust.Sci. 28 (2002) 1–74.

[10] J.F. Driscoll, Prog. Energy Combust. Sci. 34 (2008)91–134.

[11] P. Siewert, Flame Front Characteristics of TurbulentPremixed Lean Methane/Air Flames at High-pressureand High-temperature, Ph.D thesis, Swiss FederalInstitute of Technology ETH-Zurich, Zurich,Switzerland, 2005, available at <http://e-collection.ethbib.ethz.ch/ecol-pool/diss/fulltext/eth16369.pdf>.

[12] G.P. Smith, D.M. Golden, M. Frenklach, et al.,available at <http://www.me.berkeley.edu/gri_mech/>.

[13] N. Peters, J. Fluid Mech. 384 (1999) 107–132.[14] V.L. Zimont, Combust. Explo. Shock Waves 15

(1979) 305–311.[15] I.G. Shepherd, R.K. Cheng, Combust. Flame 127

(2001) 2066–2075.

[16] S. Daniele, P. Jansohn, K. Boulouchos, Flash backpropensity of syngas flames at high pressure: diag-nostic and control, Proc. ASME Turbo Expo 2010,Glasgow, UK, 2010.

[17] H. Kobayashi, K. Tamura, K. Maruta, T. Niioka,F.A. Williams, Proc. Combust. Inst. 26 (1996)389–396.

[18] H. Kobayashi, K. Seyama, H. Hagiwara, Y. Ogami,Proc. Combust. Inst. 30 (2005) 827–834.

[19] L. Landau, Acta Physicochim. Urss 19 (1944) 77–85.[20] A. Lipatnikov, J. Chomiak, Prog. Energy Combust.

Sci. 36 (2010) 1–102.[21] N. Peters, H. Wenzel, F.A. Williams, Proc. Com-

bust. Inst. 28 (2000) 235–243.[22] M. Metghalchi, J.C. Keck, Combust. Flame 38

(1980) 143–154.[23] W.T. Ashurst, G.I. Sivashinsky, V. Yakhot, Com-

bust. Sci. Technol. 62 (1988) 273–284.[24] J. Mantzaras, Combust. Sci. Technol. 86 (1992) 135–

162.[25] A.M. Klimov, Laser React. Syst. 88 (1983) 133–146.[26] V. Zimont, W. Polifke, M. Bettelini, W. Weisen-

stein, J. Eng. Gas Turb. Power-Trans. ASME 120(1998) 526–532.

[27] H. Kobayashi, Y. Kawabata, K. Maruta, Proc.Combust. Inst. 27 (1998) 941–948.