CH*/OH* CHEMILUMINESCENCE RESPONSE OF AN ATMOSPHERIC PREMIXED FLAME UNDER VARYING OPERATING CONDITIONS

Transcription

1 Proceedings of ASME Turbo Expo 1: Power for Land, Sea and Air GT1 June 1-18, 1, Glasgow, UK Proceedings of ASME Turbo Expo 1: Power for Land, Sea and Air GT1 June 1-1, 1, Glasgow, UK GT1- GT CH*/OH* CHEMILUMINESCENCE RESPONSE OF AN ATMOSPHERIC PREMIXED FLAME UNDER VARYING OPERATING CONDITIONS Daniel Guyot, Felix Guete, Bruno Scuermans Alstom CH-55 Baden Switzerland Arnaud Lacarelle, Cristian Oliver Pascereit Tecnisce Universität Berlin Germany In tis work te relationsip between te ratio of te global CH* and OH* flame cemiluminescece and te global equivalence ratio of a tecnically premixed swirl-stabilized flame is investigated. Te burner allows for a modification of te premix fuel injection pattern. Te global flame cemiluminescence is monitored by a ig-sensitivity ligt spectrometer and multiple potomultipliers. Te potomultipliers were equipped wit narrow optical band-pass filters and recorded te flame s OH*, CH* and CO* cemiluminescence intensity. To ensure an approximately uniform equivalence ratio distribution in te combustion zone, te spatial OH* and CH* flame cemiluminescence was recorded simultaneously wit one ICCD camera using a special optical setup, wic incorporated among oter tings one fully reflective and one semi-reflective mirror and appropriate optical filters. Te flame cemiluminescence intensity was mapped for a range of equivalence ratios and air mass flows. Te mapping sows tat (as stated for perfectly premixed flames in te literature) te OH*, CH* and CO* intensity of te investigated flame depends linearly on te air mass flow and exponentially on te equivalence ratio (i.e., I = km β ). Hence for te investigated operating conditions (i.e., quasi premix conditions) te global CH*/OH* intensity can be employed as a measure of te global equivalence ratio for te operating conditions investigated in tis work. However, te contribution of broadband CO* cemiluminescence in te wave lengt range of CH* cemiluminescence as to be accounted for. NOMENCLATURE Variables CW I I λ PFS R T Tr c c ligt p k B m air m f uel λ x center wavelengt cemiluminescence intensity spectral cemiluminescence intensity premix fuel split ideal gas constant temperature transmission speed of sound speed of ligt in vacuum, c ligt = 99,79,58 m s Planck constant, P = Js Boltzmann constant, k B = J K air mass flow rate fuel mass flow rate equivalence ratio wavelengt x normalized wit respect to median(x) Abbreviations PMT potomultiplier tube Subscripts x PMT x obtained from a potomultiplier tube x obtained using an optical band pass filter x BP 1 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

2 INTRODUCTION Te observation of flame cemiluminescence is a widely used tecnique in combustion diagnostics [1,, 3, ]. In particular, te intensity of OH* or CH* cemiluminescence is often used as a measure for te eat release in te flame. Detailed investigations of perfectly premixed flames ave sown tat OH or CH cemiluminescence intensity depends linearly on te air/fuel mixture mass flow rate m and exponentially on te equivalence ratio pi of te flame in te case of a globally uniform equivalence ratio distribution, i.e. I = km β, (1) were k and β are constants for one cemical species. Consequently, te ratio of bot intensity signals (i.e. CH /OH ) is a function of te equivalence ratio only. After calibration measurements, te CH*/OH* ratio provides a measure of te equivalence ratio [1]. In most applications te flame s global ligt emission is captured using one or multiple potomultipliers and fiber optic cables. From tese signals te global eat release or equivalence ratio is ten estimated. In te case of ot spots in te combustion zone due to e.g. non-uniform fuel distribution and recirculation zones etc., te potomultiplier signal migt not represent te global properties. Te relationsip between te global CH*/OH* intensity ratio and te global equivalence ratio of a steady state tecnically premixed flame in a swirl-stabilized burner is investigated. Te global flame cemiluminescence spectrum is measured and corrected for eat radiation of te combustor. Te global OH* and CH* cemiluminescence intensities are obtained from te spectrum taking into account te superposition of CH* and broadband CO* cemiluminescence. A metod is presented tat uses te spectral flame cemiluminescence results to obtain te global CH*/OH* intensity ratio of te flame from tree potomultiplier signals, wic measure te flame s OH*, CO* and superimposed CH*+CO* intensity, respectively. Te flame cemiluminescence intensity is mapped for a range of equivalence ratios and air mass flows. Te mapping sows tat (as stated for perfectly premixed flames in te literature [1, ]) te OH*, CH* and CO* intensity of te investigated flame follows Eqn. 1 (I = km β ). Te flame s global CH*/OH* intensity is sown to depend exponentially on te equivalence ratio and to be independent of te air mass flow for te operating conditions investigated in tis work. Hence te global CH*/OH* intensity can be employed as a measure of te global equivalence ratio. Altoug only steady flames were investigated in tis work, te identified relationsip between equivalence ratio and CH*/OH* intensity ratio sould also be valid for unsteady flames, provided tat a uniform spatial distribution of te equivalence ratio is maintained at all times. Te work presented ere is a continuation of te work presented in [5], were te flame transfer function of an industrial swirl-stabilized burner under full engine pressure was determined from global flame cemiluminescence recording. EXPERIMENTAL SET-UP Combustion Facility All combustion results presented in tis paper were obtained using te combustion facility sown in Fig. 1. On te upstream side te combustion test facility features an air preeater troug wic te combustion air is fed into a duct. Inside tis duct sits te burner lance, troug wic fuel (natural gas in tis work) is supplied to te burner. Te atmosperic combustor as a 3 mm long air-cooled quartz glass combustion camber allowing for optical access to te flame. A water-cooled resonance tube of 135 mm in lengt is attaced to tis combustion camber. Te resonance tube consists of tree parts mounted togeter by flanges. Te upstream duct and te middle part of te resonance tube are equipped wit arrays of water-cooled micropone olders. One speaker on te upstream end and two speakers on te downstream end of te test facility allow for acoustic excitation. EV-1 Burner Te combustor incorporates a generic environmental burner (EV-1) designed by Alstom wit a cross-sectional area expansion ratio of : 1 for flame stabilization. Figure sows a detailed sketc of te burner. It is composed of two alf cones sifted in suc a way tat te air is forced to enter te cone circumferentially troug two slots. Te resulting swirling air flow generates a recirculation zone along te centerline at te burner outlet, tus stabilizing te flame in tis region. In te standard configuration te main (premix) fuel is injected troug 6 boreoles,.7 mm in diameter eac, wic are distributed equidistantly along te burner s two air slots and fed from one common fuel supply. Mixing of swirling air and main fuel results in quasi premixed combustion. To enable control of te fuel distribution profile and ence te flame location, an additional premix fuel injector was installed in te burner slots. Tis injector was divided into two sections as illustrated in Fig. : an upstream section wit 1 boreoles and a downstream section wit 16 boreoles. Te fuel supply to te two upstream sections and te two downstream sections can be Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

3 individually controlled by two mass flow controllers. Tis configuration allows control of te premix fuel split PFS, wic is defined ere as te ratio of premix fuel mass flow troug te upstream injector section m premix,us to te total premix fuel mass flow troug upstream & downstream section m premix,tot : PFS = m premix,us m premix,tot. () potomultiplier OH * (CW 38 nm) CO * (CW 7 nm) CH * & CO * (CW 31 nm) CO * (CW 51 nm) Figure 3 sows images of te flame for different values of te PFS wit oterwise identical operating conditions ( =.6,m air = kg/,t inlet = 15 C). Te effect of te PFS can be observed clearly: For PFS = 1 all premix fuel is injected troug te upstream injector sections. For tis case te fuel mainly ends up in te inner recirculation zone and te flame reaces deep down into te burner cone. For PFS = all premix fuel is injected troug te downstream sections. More fuel ends up in te outer recirculation zone and te flame stabilizes downstream of te burner. At PFS =.3 (i.e., sligtly more fuel flow troug te downstream injector sections) te flame stabilizes just downstream of te burner and sows a very uniform flame cemiluminescence intensity distribution. ligt spektrometer ( 8 nm) PFS = PFS =.3 PFS =.5 PFS = 1. Figure 3. Effect of premix fuel split on te flame location. Figure 1. Combustion test facility. Aside from premix fuel injection, pilot fuel can be injected at te EV-1 cone apex using a pilot lance. However, only operating conditions witout pilot fuel injection were investigated in tis work. For a detailed description of te burner see Ref. 6. additional injectors upstream downstream standard injectors air flow Figure. premix fuel additional injectors standard injectors Modified Alstom EV-1 burner. outer recirculation zone inner recirculation zone sear layer Sensors Te spatial distribution of OH* and CH* cemiluminescence in te flame was recorded simultaneously by one single ICCD camera. To allow for te simultaneous recoring a special optical setup was used (Fig., wicincorporated an optical lens, two mirrors (one fully reflective, one semi reflective) and two optical filters. Te semi reflective mirror was used to split te ligt emitted by te flame into two images, wic were ten transmitted troug two narrow band-pass filters centered at 38 nm and 31 nm, respectively. By positioning lens, mirrors, filters and te ICCD camera appropriately te ICCD sensor was recording two flame images, one representing te flame s spacial OH* cemiluminescence, te oter one te flame s spacial CH* cemiluminescence (including te CO broad-band). Te transmission of te optical filters is sown in Fig. 5. In addition to te ICCD recordings, te ligt emitted 3 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

4 ICCD OH* CH* CW 38nm CW 31nm optical band-pass filters fully reflective semi reflective mirrors lens flame dividual filters is sown in Fig. 5. Te micropone and potomultiplier signals were amplified and low-pass filtered at khz to avoid aliasing. Te ig-sensitivity ligt spectrometer was used to measure te ligt spectrum of te flame. A standard poto camera recorded images of te flame witin te visible area of ligt. Pressure oscillations in te combustion camber were measured using a condenser micropone placed into te first micropone older downstream of te flame. Figure. Optical setup for simultaneous recording of te spatial distribution of OH* and CH* cemiluminescence in te flame. Tr λ in % Figure 5. CW 38 nm (OH*) CW 7 nm (CO*) CW 31 nm (CH*) CW 51 nm (CO*) λ in nm Transmission of te optical band-pass filters. by te flame was also captured by a fiber optic probe (see Fig. 1). At te optical probe s end tere is a 1.8 mm tick optical fiber directed at te flame. Te ligt collected by tis tick fiber was passed on to 1 very small fibers forming a bundle of also 1.8 mm tickness via an optical splitter. Te 1 small fibers are bundled togeter in groups of 3 to form 7 fiber bundles to wic potomultiplier tubes or oter optical sensors can be connected. Tis rater complicated set-up of te optical probe ensures tat eac of te connected optical instruments receives te same ligt from te flame. However, for tis work only five of tese seven bundles were used to connect four potomultiplier tubes and a ig-sensitivity ligt spectrometer to te probe. Te potomultipliers were equipped wit narrow bandpass filters centered at 38, 7, 31, and 51 nm. At tese wavelengts, tey captured ligt from OH,teCO broad-band, CH (including te CO broad-band), and again only te CO broad-band cemiluminescence, respectively. Te optical widt of te eac optical filters smaller tan +-1 nm to prevent overlapping of te bandwidt range of transmission. Te transmission of te in- Te Flame Cemiluminescence Spectrum Measured Ligt Spectrum For eac burner operating condition te ligt emission of te flame was measured using an Ocean Optics QE65 ig-sensitivity ligt spectrometer. Te spectrometer incorporates a Hamamatsu back-tinned detector tat is responsive from -11nm, features low readout noise, and can be cooled wit an onboard termoelectric cooler to reduce dark noise. As an example, Fig. 6 sows te measured ligt spectrum obtained for te operating conditions, wic will later be used as te reference conditions (black line). Note tat te spectrometer sensitivity and te transmission of te optical probe ave been accounted for in te presented spectrum. Te distinct peaks corresponding to OH and CH cemiluminescence are clearly visible, as is te more distributed CO cemiluminescence. At 51 nm a small peak attributed to C cemiluminescence can also be seen. I λ (normalized) λ in nm Figure 6. Flame cemiluminescence spectrum: spectrum obtained from te ligt spectrometer (black), fitted black body spectrum (red), and corrected flame spectrum (blue). Te black dotted lines indicate te CW of te employed optical filter. Operating conditions: =.65, m air = kg/,t inlet = 15C. Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

5 Black Body Correction Additionally, a steep intensity increase is present above 6 nm. Tis increase is due to eat radiation of te burner plate, wic was witin te field of view of te optical probe. To obtain only te flame cemiluminescence spectrum, a black body radiation spectrum was fitted to te measured spectrum above 67 nm. For tis fit te radiation intensity of a black body was assumed to follow Planck s Law: I λ,bb (λ,t BB )=A Pc ligt λ 5 ( P c ligt 1 e λk B T BB 1), (3) were I λ,bb is te spectral black body radiation, T BB te blackbodytemperature,a a scaling constant c ligt te speed of ligt, P te Planck constant, and k B te Boltzmann constant. Te flame spectrum (Fig. 6, blue line) was ten obtained by subtracting te approximated radiation spectrum (Fig. 6, red line) from te measured spectrum. In te fitting routine te two unknowns, te scaling constant A and te black body temperature T BB, were fitted to te flame spectrum. Wile te scaling constant A was very similar for different operating conditions, te approximated black body temperature T BB increased (as expected) wit te eat release in te flame (i.e., wit increasing fuel mass flow), indicating a otter burner plate. For te flame spectrum presented in Fig. 6 T BB was approximated to be 15 K, wic is in te order of te expected surface temperature of te burner plate, tus giving confidence into te correction approac. Subtraction of te CO Contribution in te Flame Spectrum In tis work te equivalence ratio of te flame is related to te ratio of CH to OH cemiluminescence intensity. Since in te flame spectrum te CH cemiluminescence peak is superimposed wit te broadband CO emission, te CO contribution as to be subtracted to obtain only te CH emission [7, 8]. To do so, te CO emission spectrum is obtained in a second fitting routine. Te CO intensity is assumed to follow an extreme fit function as proposed by Seipel et al. [7]: [ ( (λ λc ) I λ,co = A exp exp w ) (λ λ ] C) + 1, () w were I λ,co is te normalized spectral CO cemiluminescence intensity, A a scaling constant, λ te wavelengt, λ C te wavelengt of maximum intensity, and w a scaling constant (in nm) for te widt of te extremum. Hence, te sape of te intensity distribution over te wavelengt is only dependent on te two variables λ C and w. Te fit was performed in te wavelengt ranges around te CH peak, were only CO emission is present in te spectrum. Figure 7 presents te flame spectrum togeter wit te fitted CO contribution and te flame spectrum witout te CO contribution for te reference operating point. Te fit agrees very well wit te background found inteflamespectrum. I λ (normalized) λ in nm Figure 7. Flame cemiluminescence spectrum wit extreme fit for te CO emission: flame spectrum (blue), CO fit (green), and flame spectrum witout CO contribution (magenta). Operating conditions: =.65, m air = kg/,t inlet = 15C. By subtracting te approximated CO contribution from te flame spectrum te CH peak is isolated at approximately 31 nm. By integrating te spectral CH intensity over te wavelengt te absolute CH intensity can be computed: ( ) I CH = I λ,ch dλ = Iλ, f lame I λ,co dλ, (5) λ CH λ CH were λ CH indicates te wavelengt range in wic CH cemiluminescence occurs. IntesamewayteabsoluteOH intensity can be obtained by integrating te spectral OH intensity. Note, owever, tat since te CO is approximately zero witin te wavelengt range of OH cemiluminescence, no correction for CO emission is required: I OH = I λ,oh dλ. (6) λ OH Figures 8 and 9 present te cange of cemiluminescence intensity wit air mass flow and equivalence ratio. 5 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

6 intensity λ in nm Figure 8. Spectral intensity vs. air mass flow (=.65, T inlet =15 ). emission compared to, for example, diffusion flames. For most ligt spectra presented in tis work te recording time was 1 s to fully utilize te buffer limits of te ligt detector, altoug good results were also obtained wit recording times of 1 s. If te recording time is to be reduced even furter, averaging of multiple recordings is a possible way to maintain a good signal-to-noise ratio. To still resolve fluctuations using averaging a pase-logged triggering of te spectrometer would ave been required, wic would ave dramatically increased te necessary measurement time. Terefore, te CH cemiluminescence intensity ad to be obtained using potomultiplier tubes, wic offer a ig gain, low noise and ig frequency response. Subtraction of te CO Contribution in te PMT Signals Te output voltage U PMT obtained from a potomultiplier can be expressed as: intensity λ in nm Figure 9. Spectral intensity vs. air mass flow (m air =, T inlet =15 ). Wit te outlined approac te computation of te OH and CH cemiluminescence intensity from te ligt spectrometer data is straigtforward for a steady flame spectrum. In te case of eat release oscillations (forced or self-induced), owever, te flame spectrum is not going to be steady. For te frequency range of interest for combustion oscillations (i.e., 3 to 5 Hz), tis is generally not possible wit commercial ligt spectrometers. Te igsensitivity ligt spectrometer used in tis work, for example, as a minimal sutter speed of 8 ms, wic would allow frequencies of up to 6.5 Hz to be resolved (not accounting for readout time). Furtermore, te recording times needed to acieve a good signal-to-noise ratio are commonly muc longer, especially since premix flames feature only low ligt (( ) ) U PMT = g PMT Iλ, f lame + I λ,bb Trλ,probe Tr λ,bp Se λ,pmt dλ +U PMT,dark (7), were Tr λ,probe and Tr λ,bp denote te spectral transmission of te fiber optic probe and te optical bandpass filter, respectively, and Se λ,pmt te spectral sensitivity of te PMT. g PMT represents te PMT gain, wic was old constant for eac PMT trougout te wole test campaign. Te dark signal U PMT,dark of all PMTs was adjusted to zero before te first measurement by applying an offset voltage. Since te probe transmission and te PMT sensitivity are approximately constant witin te narrow wavelengt range of transmission of te employed optical band pass filters, tey can be combined wit te gain g PMT to a new gain G PMT. Also, te black body radiation is approximately zero witin te range of te band pass filters: G PMT = g PMT λ BP ( Trλ,probe Se λ,pmt ) dλ, (8) U PMT = G PMT λ BP ( Iλ, f lame Tr λ,bp ) dλ, (9) were λ BP denotes te band pass filter s wavelengt range of transmission, i.e., approximately 99 to 318 nm in case of te CW 38 nm band pass filter (99 nm < λ CW 38 < 318 nm). As a measure for te OH cemiluminescence intensity te output voltage of te PMT equipped wit te 6 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

7 CW 38 nm band pass filter was used. As stated earlier, te signal does not ave to be corrected for CO,sinceno CO emission is present witin te wavelengt range of tis optical filter. To obtain a signal tat only represents te CH intensity, owever, a combination of ligt spectrometer and PMT measurements wit a steady flame is required to account for te CO emission captured by te CW 31 nm filter. According to Eq. (9) te output voltage of te PMT CW 31 can be expressed as: U CW 31 = U CH,CW 31 +U CO,CW 31 = G CW 31 λ CW 31 ( Iλ,CH + I λ,co ) Trλ,CW 31 dλ, (1) were U CH,CW 31 and U CO,CW 31 represent te contribution of CH* and CO* to te total output voltage of PMT CW31. Since te (normalized) spectral intensities, te filter transmission, and te output voltage are known, te gain G CW 31 can be determined. Once te gain is known te output voltage U CW 31 can be separated into its CH and CO contribution: U CH,CW 31 = G CW 31 λ CW 31 I λ,ch Tr λ,cw 31 dλ (11) U CO,CW 31 = G CW 31 λ CW 31 I λ,co Tr λ,cw 31 dλ. (1) Since te CO intensities acquired at center wavelengts of 7, 31 and 51 nm are proportional to eac oter, U CO,CW 31 can be related to te output voltages acquired at te oter two wavelengt ranges by two proportionality constants C CO,CW 7 and C CO,CW 51 : U CO,CW 31 = C CO,CW 7 U CW 7 (13) U CO,CW 31 = C CO,CW 51 U CW 51. (1) Te proportionality constants do not only relate te PMT output voltages to eac oter, but also te integral cemiluminescence intensities, wic can be computed from te spectrometer data: I CO,CW 31 = C CO,CW 7 I CO,CW 7 (15) I CO,CW 31 = C CO,CW 51 I CO,CW 51, (16) were te integral CO* intensity for eac band pass filter is given by I CO,BP = λ BP I λ,co Tr λ,bp dλ. Te spectral CO* intensity and te filter transmissions are known, te proportionality constants can been determined, and U CO,CW 31 can be obtained from eiter te U CW 7 or te U CW 51 PMT signal. In tis work te CH signal was determined using te average of te CO contributions at 31 nm obtained from Eqs. (13) and(1). Note tat te proportionality constants are also valid for an unsteady flame wit cemiluminescence intensity fluctuations. Hence, te described approac is not limited to steady flames and te CO contribution in an unsteady flame and ence unsteady U CW 31 signal could be subtracted to obtain U CH,CW 31. SPATIAL CHEMILUMINESCENCE INTENSITY SIGNALS Using te ratio of global OH* and CH* intensity of a flame as a measure of te flame s global eat release is only exact, if te flame features a uniform spacial equivalence ratio distribution across te flame front. Tis is due to te non-linear relationsip between te flame cemiluminescence intensity and te equivalence ratio. For perfectly premixed flames te spacial equivalence ratio distribution across te flame front is uniform per definition. A tecnically premixed flame like te one investigated in tis work is typically not perfectly premixed, but can be assumed to be quasi premixed. If te standard deviation of te spacial equivalence ratio distribution is only small, te use of global cemiluminescence measurements is still justified. To assess te flame s spacial equivalence ratio distribution te spacial OH* and CH* intensity as been recorded simultaneously on one ICCD camera detector using te setup sketced in Fig.. Te recording was performed for =.6,m air = kg/,t inlet = 15 C and a PFS of.3. Figure 1 sows te obtained OH* and CH* intensity distribution. A gray color map is used wit wite indication te maximum intensity and darker colors indication lower intensity. All points wit an intensity lower ten 15 % of te maximum intensity are plotted in black to igligt te contour of te flame. A clear spacial distribution of te cemiluminescence intensities is visible. Te OH* and CH* intensity distribution is very similar. Note tat te CH* intensity as measured by te ICCD camera also includes te cemiluminescence contribution of CO*. From te spacial distribution of OH* and CH* intensity te spacial CH*/OH* ratio is computed. To illustrate te distribution of te spacial CH*/OH* ratio, te relative deviation of te local to te average CH*/OH* ratio is presented in Fig. 11. Te figure sows only small deviations 7 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

8 OH* CH* Figure 1. Simultaneous OH* (left) and CH* (rigt) intensity distribution as recorded by te ICCD camera ( =.6,m air = kg/,t inlet = 15 C, PFS of.3). constant at 15 C. At operating points witin tis parameter range, te time traces of te flame s cemiluminescence were recorded by te potomultipliers. In addition, images of te flame were taken using a standard poto camera. Note tat wen determining te coefficient beta in Eq. (17) te scaling factor k are also determined. Figure 1 sows te recorded images for selected combinations of air mass flow and equivalence ratios. In tese images, te mean flow direction is from bottom to top. Te burner plate appears on te bottom as a red glowing disc wit te burner exit in te center. Te flame is stabilized downstream of te burner exit, were its ligt emission is completely captured by te optical probe. between te local and te average CH*/OH* ratio. Most areas of te flame ave a deviation of - 5 % (wite) and te igest deviation is in te range of 15 - % in only small areas. Tis indicates an approximately uniform equivalence distribution across te flame front. Hence, te use of global cemiluminecence measurements appears justified for tis operating conditions. For premix fuel splits of.5 and 1 te spacial CH*/OH* ratio distribution was less uniform. Terefore, a PFS of.3 was selected for te investigations of te global cemiluminescence response. CH*/OH* distribution Deviation of local CH*/OH* from average CH*/OH* - 5% 5-1% 1-15% 15 - % > % Figure 11. CH*/OH* ratio distribution as deviation between te local and te average CH*/OH* ratio. m air = 18 kg/ m air = kg/ m air = kg/ =.6 =.65 =.75 Figure 1. Poto images of te flame at different air mass flows and equivalence ratios (T inlet =15 ). GLOBAL CHEMILUMINESCENCE INTENSITY SIGNALS Calibration of te Potomultiplier Signals To obtain te coefficients β OH, β CH,andβ CO in I = km β, (17) te EV-1 burner was operated at different combustion air mass flow rates (17 to 3 kg/) and equivalence ratios (.55 to.8), wile te burner inlet temperature was eld After te calibration measurements ad been performed, te constants k and β were determined by fitting Eq. (17) to te average cemiluminescence intensities of OH,CH and CO. Figure 13 sows te recorded cemiluminescence intensity obtained directly from te four potomultipliers togeter wit te corresponding intensity fits. Te recorded data points are indicated by circles. Te fit is presented as a surface plot. Te color of te data point circles indicates te deviation between fit and measurement normalized by 8 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

9 te intensity range of te measurement points. Black circles indicates an error of less tan %, magenta circles an error between to 8%. Te obtained coefficients k and β are given in Tab. 1 togeter wit te coefficients of determination R between eac fit and te respective measured values. Table 1. Coefficients k and β obtained from te calibration measurements and coefficients of determination R between fit and measured values. OH CO CH &CO CH CO CO CW 38 nm CW 7 nm CW 31 nm CW 31 nm CW 31 nm CW 51 nm k β R UPMT in V 18 OH* (CW 38 nm) CO* (CW 7 nm).8 As can be seen in Fig. 13, te trends found in te measured cemiluminescence intensity are well captured by te fits wit relative errors between fit and measurement mainly below %. Tis is also evident from te R -values in Tab.1, wic are all very close to 1. Te lowest R -value (.9) was obtained for te fit of te CW 31 nm signal, because te corresponding signal is a superposition of CH* and CO* intensity. Since te cemiluminescence intensity of eac species depends differently on te equivalence ratio (i.e., β CH β CO ), te combined cemiluminescence intensity can not be captured perfectly by a fitting function I = km air β, altoug te matc is still good. For all oter signals, te R -values is.97 or larger, wic indicates a very good fit. As previously mentioned, te PMTs capture only CO* at CW 7 nm and CW 51 nm. Terefore, te dependence of te two PMT signals on air mass flow and equivalence ratio sould be te same (i.e., β CW 7 = β CW 51 ). Indeed, te two β coefficients are approximately equal. In contrast, k CW 7 and k CW 51 cannot be expected to be equal since tey depend on te respective absolute CO* intensity, te filter transmission and potomultiplier gain. To illustrate te quality of te OH intensity fit in more detail, Fig. 15 presents te measured OH intensity as a function of te air mass flow and equivalence ratio togeter wit lines of constant air mass flow and equivalence ratio according to te fit. Te fit captures te measured trends very well. Te results confirm te linear dependence of te OH intensity on air mass flow and te exponential dependence on equivalence ratio. Te measured intensities of CH and CO cemiluminescence (not sown) also matc well wit teir fits. To estimate te CH* intensity signal, te proportionality constants C CO,CW 7 and C CO,CW 51 and te gain G CW 51 UPMT in V UPMT in V UPMT in V CH* & CO* (CW 31 nm) CO* (CW 51 nm) Figure 13. Recorded intensities of OH,CH and CO cemiluminescence measured by te potomultipliers and corresponding surface fits as a function of air mass flow and equivalence ratio. 9 Copyrigt 1 by ASME and Alstom Tecnology, Ltd

10 Figure 1. UCH,CW 31 in V UCO,CW 31 in V CH* (CW 31 nm) CO* (CW 31 nm) Approximated intensities of OH,CH and CO cemiluminescence measured by te potomultipliers and corresponding surface fits as a function of air mass flow and equivalence ratio. UCW 38 in V UCW 38 in V =.55 (fit) =.6 (fit) =.65 (fit) =.7 (fit) =.75 (fit) =.8 (fit) measurement m air = 17 kg/ (fit) m air = kg/ (fit) m = 3 kg/ (fit) air measurement Figure 15. Recorded intensity of OH cemiluminescence and lines of constant air mass flow (top) and equivalence ratio (bottom) were computed for eac calibration measurement point according to Eqs. (15), (16) and(9), respectively. Te results are presented in Fig. 16. Te actual values are normalized by teir median (indicated by te tilde) and plotted against te equivalence ratio. For te proposed approac to determine te CH* intensity from te PMT signals, te proportionality constants and te gain ave to be constant trougout te wole range of air mass flows and equivalence ratios investigated. Indeed, tey are approximately constant for all calibration points wit standard deviations of less ten %. Note, tat for furter computations te median of te proportionality constants was used for all burner operating conditions. CCO,CW 7 CCO,CW 51 GCW Figure 16. Distribution of te proportionality constants C CO (top, middle) and te gain G (bottom) needed to obtain te CH intensity from te PMT signals. Values are normalized wit respect to te median of all calibration measurement points. Using Eqs. (1), (11) and (1) te CH* and CO* contribution to U CW 31 can ten be identified. By fitting Eq. (17) (I = km air β ) to te CH* and CO* intensity, te corresponding constants k and β were determined in te same way as for te potomultiplier signals. Te intensity data and te fit are sown in Fig. 13 (bottom), wile te constants k and β are given in Tab. 1. As for te oter signals te fit captures te data trends very well. Since te CO* intensity at CW 31 nm was approximated using te CW 7 nm and CW 51 nm intensities, β CO,CW 31 is identical to β CW 7 and β CW 51. Additionally, te R -value of te CO* and CH* fits for CW 31 nm are somewat closer to 1 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

11 1 tan for te fit of te total PMT voltage. Tis indicates tat te measured intensity data better represents te combination of two intensity functions, wic furter increases te validity of te used separation approac. After te CH* intensity as been determined for eac calibration point, te I CH I OH ratio can be calculated. Figure 17 presents tis ratio as a function of te equivalence ratio (bottom). Te results for different air mass flows are indicated by different symbols and colors. For comparison, te ratio of PMT signals U CW 31 U CW 38 (i.e., no correction for CO*) are plotted as well (top). Wile bot plots sow te typical increase of te I CH I OH ratio wit equivalence ratio, te ratio of te uncorrected PMT signals sows sligtly sifted trends for different air mass flows, due to te effect of CO*. Wen using te isolated CH* signal, owever, te results from all investigated air mass flows fall on one single trend line. Tis trend line can be directly obtained from te fitting constants k and β for te OH* and CH* intensity: I CH I OH = k CHm air βch k OH m air β OH = k CH k OH (β CH β OH ) (18) Because te I CH I OH ratio is independent of te air mass flow, it can be used to approximate te equivalence ratio of te flame from optical signals only. SUMMARY & CONCLUSIONS Te relationsip between te global CH*/OH* intensity ratio and te global equivalence ratio of a tecnically premixed flame in a swirl-stabilized burner as been investigated. For tis investigation a specific premix fuel gas injection pattern (PFS =.3) was selected, since for tis PFS te flame featured an approximately uniform spatial CH*/OH* intensity ratio distribution (and ence an approximately uniform spatial equivalence ratio distribution) and stable combustion. Te global flame cemiluminescence spectrum was measured and corrected for eat radiation of te combustor. Te global OH* and CH* cemiluminescence intensities were obtained from te spectrum taking into account te superposition of CH* and broadband CO* cemiluminescence. A metod was developed tat uses te spectral flame cemiluminescence results to obtain te global CH*/OH* intensity ratio of te flame from tree potomultiplier signals, wic measure te flame s OH*, CO* and superimposed CH*+CO* intensity, respectively. For increased accuracy a fourt potomultiplier was employed to obtain a second CO* intensity signal. UCW 31 UCW 38 UCH,CW 31 UCW fit m air = 17 kg m air = 18 kg m air = 19 kg m air = kg m air = 1 kg m air = kg m air = 3 kg Figure 17. Variation of te output voltage ratio of PMT CW31 and PMT CW38 (top, no correction for CO*) and te ratio of CH* and OH* intensity (bottom, CH* corrected for CO*) as obtained from te PMTs wit air flow and equivalence ratio. Te used fit function for te I CH I ratio OH is k CH k OH (β CH β OH ) wit values of k and β according to Tab. 1. Te flame cemiluminescence intensity was mapped for a range of equivalence ratios and air mass flows. Te mapping sows tat (as stated for perfectly premixed flames in te literature) te OH*, CH* and CO* intensity of te investigated flame follows Eqn. 1 (I = km β ). Te flame s global CH*/OH* intensity was sown to depend exponentially on te equivalence ratio and to be independent of te air mass flow. Hence te global CH*/OH* intensity can be employed as a measure of te global equivalence ratio for te operating conditions investigated in tis work. However, te contribution of broadband CO* cemiluminescence in te wave lengt range of CH* cemilumi- 11 Copyrigt 1 by ASME and Alstom Tecnology, Ltd.

12 nescence as to be accounted for. Note tat tis can only be done wit at least tree potomultipliers and an additional measurement of te flame s cemiluminescence spectrum and only works for quasi premix conditions in te flame front. Altoug only steady flames were investigated in tis work, te identified relationsip between equivalence ratio and CH*/OH* intensity ratio sould also be valid for unsteady flames, provided tat a uniform spatial distribution of te equivalence ratio is maintained at all times. [7] Seipel, A., Brockinke, A., and Kose-Hoingaus, K., 9. Tp3: Spectroscopic caracterization and simulation of cemiluminescence. nd International Worksop on Cemiluminescence and Heat Release, Müncen. [8] Lauer, M., and Sattelmayer, T., 9. On te adequacy of cemiluminescence as a measure for eat release in turbulent flames wit mixture gradients. ASME GT , Proc. ASME Turbo Expo 9, Orlando, June 8-1. ACKNOWLEDGMENT Te investigations were conducted as part of te joint researc program COOREFF-T in te frame of AG Turbo. Te work was supported by te Bundesministerium für Wirtscaft und Tecnologie (BMWi) undergrant number 3771J. Te autors gratefully acknowledge AG Turbo for its support and permission to publis tis paper. Te responsibility for te content lies solely wit its autors. REFERENCES [1] Higgins, B., McQuay, M., Lacas, F., Rolon, J., Darabia, N., and Candel, S., 1. Systematic measurements of o cemiluminescence for fuel-lean igpressure, premixed, laminar flames. FUEL 8, 1, PP [] Higgins, B., McQuay, M., Lacas, F., and Candel, S., 1. An experimantal study of pressure and strain rate on c cemiluminescence on premixed fuel-lean metans /air flames. FUEL 8, 1, PP [3] Nori, V., and Seitzman, J., 7. Detailed distributions of o*, c* and c* cemiluminescence in te reaction zone of laminar metane/air premixed flames. AIAA-7-66 at te 5t Aerospace Sciences Meeting, Reno, NV, Jan 8-11, 7. [] Kojima, J., Ikeda, Y., and Nakajima, T.,. Detailed distributions of o*, c* and c* cemiluminescence in te reaction zone of laminar metane/air premixed flames. 36t AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exibit () AIAA-339. [5] Scuermans, B., Guete, F., Pennell, D., Guyot, D., and Pascereit, C. O., 9. Termoacoustic modeling of a gas turbine using transfer functions measured at full engine pressure. ASME Paper GT9-5965, Proc. ASME Turbo Expo 9, Orlando, June 8-1. [6] Döbbeling, K., Knöpfel, H. P., Polifke, W., Winkler, D., Steinbac, C., and Sattelmayer, T., 199. Low NOx Premixed Combustion of MBtu Fuels Using te ABB Double Cone Burner (EV Burner). ASME Paper 9- GT Copyrigt 1 by ASME and Alstom Tecnology, Ltd.