Influence of Fuel Staging on Thermo-Acoustic Oscillations in a Premixed Stratified Dual-Swirl Gas Turbine Model Combustor

In several technical combustion systems for lean, premixed combustion, mixture stratification plays an important role, such as in stationary and aero gas turbines. The current paper focuses on a detailed characterization of a dual-swirl gas turbine model combustor operated in a stratified regime. The influence of mixture stratification on flame stabilization and self-induced thermo-acoustic oscillations was studied using laser and optical diagnostics in combination with microphone probes to measure pressure oscillations in the plenums and combustion chamber. The overall flame-shape was imaged using OH* chemiluminescence. Laser Raman scattering was applied to study the thermo-chemical state of the flame, the flow-field was measured using Particle Image Velocimetry and pressure oscillations in the combustion chamber and in the air plenums were determined by using calibrated microphone probes. OH planar laser-induced fluorescence was used to determine fluctuations of the hot gas with high spatial resolution. Significant mixture stratification within the combustion chamber was found to only occur at or upstream of the flame root. Varying the stratification level did not significantly influence the mean flame shape and flow field, however, a clear influence on thermo-acoustic oscillations was observed.


Introduction
In gas turbines for land based [1] and aeronautical [2] applications, lean premixed combustion is becoming of increased importance due to the potential for the reduction of pollutants. However, in practical applications, mixture stratification (i.e. non-uniform mixture fraction distributions) can occur, e.g. due to using a pilot stage or due to imperfect mixing, thus influencing the flame stabilization and flame dynamics.
Stratified flames in generic configurations have gained significant research interest in recent years, both experimentally and numerically. For example Sweeney et al. applied laser-based measurements of flow field and species concentration in stratified low turbulence flames in a slot burner [3] and in non-swirling and swirling stratified flames in a double-annular bluff body burner [4,5]. They found significant levels of stratification within the flame front. Seffrin et al. and Kuenne et al. studied a series of flames in a coannular burner with a flame holder and well-defined boundary conditions, the so-called Darmstadt stratified flame [6,7]. They discovered considerable influence of the operating conditions on the flame topology. Meares et al. [8,9] studied the effect of different levels of mixture fraction inhomogeneities in a co-annular tube burner. Here, mixture inhomogeneities significantly influenced the stabilization limit of the burner.
Several of these configurations have also been studied numerically, and are target flames of the International Workshop on Turbulent Flames (TNF) [10]. A recent overview of experiments and simulations of stratified flames is given in the review paper by Masri [11]. Besides the above mentioned simple stratified flame configurations, several other model combustors with technically premixed flames, resulting in flames with mixture stratification, are described in the literature (e.g. [12][13][14]).
In the current study, stratified flames are investigated in the SFB 606 dual-swirl gas turbine model combustor (GTMC), which provides a technically relevant geometry (i.e. swirl, high turbulence levels and thermal powers, enclosed flame). The burner features two concentric co-rotating swirlers, each supplied by a separate plenum, allowing to feed each of the two concentric flows with premixed mixtures of different equivalence ratios. This GTMC has been studied previously with focus on technically premixed methane flames [15][16][17] and prevaporized liquid fuel flames [18,19]. A nearly identical setup is available for liquid fuels [20]. Further, the design allows for several identical burners to be combined into a combustor array, in order to investigate cross-flame interactions [21].
In the current paper, the focus is on a detailed characterization of the influence of stratification on selfexcited thermo-acoustic oscillations. Three flames, one perfectly premixed flame and two flames with different stratification levels, were studied using laser-based and optical measurement techniques. The flow-fields were characterized by stereoscopic particle image velocimetry (PIV), the flame shapes were visualized by planar laser-induced fluorescence (PLIF) of OH and OH * chemiluminescence imaging, and laser Raman scattering was used to determine the thermo-chemical state of the flames. In addition, microphone probes in the two plenum chambers and in the combustion chamber allowed the assessment of the acoustic boundary conditions. Different acoustic modes of the combustor were excited by the flames and the level of excitation changed with the mixture stratification. The results demonstrate how thermo-acoustic instabilities in gas turbine flames can be influenced by modifying and controlling the mixture stratification.

SFB 606 Gas Turbine Model Combustor
A schematic of the combustor is shown in Figure 1. Details on the geometry can be found in the literature [15], and only a brief overview is provided here. The combustor has two concentric, co-rotating swirlers, each being supplied through a separate plenum chamber. Thus, the gas supply to each plenum (and swirler) can be set independently, allowing control of the air split ratio L (defined as the ratio of the mass flows of the outer and inner plenum). For the current study, all flames were operated at an air split ratio L = 1.6, corresponding to equal pressure loss across both swirlers. Different fuel/air mixtures were supplied to each swirler by premixing fuel and air far upstream of the plenum inlets using static mixers (Sulzer). The vertically arranged combustion chamber offers very good optical access from all four sides and is equipped with several ports for dynamic pressure probes. Table 1 shows the operating conditions. All studied flames had a global equivalence ratio of φglobal = 0.75 and a thermal power of Pth = 25 kW. The mixture fraction in the inner and outer plenum was varied, resulting in different stratification ratios S between 1 (corresponding to a perfectly premixed flame) and 1.67. Here, S is defined as the ratio of the equivalence ratio in the inner (φin) and outer (φout) stream.

OH Planar Laser-Induced Fluorescence (PLIF) and OH * Chemiluminescence (CL) Imaging
OH planar laser-induced fluorescence was measured at a frame rate of 10 kHz using a Sirah Credo dye laser pumped by a diode-pumped solid state (DPSS) laser (Edgewave Innoslab IS400-2-L). The frequency doubled output of the dye laser was tuned to 283.2 nm to match the Q1(7) transition of OH in the A-X (ν'' = 1, ν' = 0) band. The laser beam was expanded into a light sheet using a two-stage cylindrical telescope and focused into the test section using a third cylindrical lens, resulting in a laser sheet of approximately 100 mm height with a beam waist in the test section of approximately 0.5 mm.
The fluorescence signal in the (1,1) and (0,0) band around 308 nm was collected using a high speed OH * chemiluminescence was simultaneously imaged using the same camera / image intensifier combination used for OH PLIF. Here, the camera was operated at a frame rate of 20 kHz with alternating short (300 ns for OH PLIF) and long (15 µs for OH * CL) image intensifier gates, resulting in OH PLIF images for the even camera frames and OH * CL images for the odd camera frames. 30,000 single shot image pairs were recorded for each operating condition. The corresponding trigger scheme for the camera, image intensifier (I/I) and laser is shown in Figure 2. The benefit of this setup lies in the reduced optical access necessary for the two measurement quantities and a decreased experimental complexity and cost as well as a perfect, calibration-free, spatial overlap between OH PLIF and OH * CL.

Stereoscopic Particle Image Velocimetry (S-PIV)
The

Laser Raman Scattering
Single-shot laser Raman scattering was applied for the pointwise quantitative measurement of the major species concentrations (O2, N2, CH4, H2, CO, CO2, H2O) and the temperature. Raman measurements were only performed for the operating condition with the highest stratification ratio (S = 1.67), as the highest mixture fraction variations are expected here. Details of the measurement system can be found in [14,22] and only a short summary is given here. The radiation of a flashlamp-pumped dye laser (Candela LFDL 20, 1.8 J/pulse at 489 nm, pulse duration 3 µs, pulse repetition rate 5 Hz) was focused into the combustion chamber and the Raman scattering emitted from the measurement volume (length ≈ 0.6 mm, diameter ≈ 0.6 mm) was collected by an achromatic lens (D = 80 mm, f = 160 mm) and relayed to the entrance slit of a spectrograph (SPEX 1802, f = 1 m, slit width 2 mm, dispersion ≈ 0.5 nm/mm). The dispersed and spatially separated signals from the different species were detected by individual photomultiplier tubes (PMTs) in the focal plane of the spectrograph and sampled using boxcar integrators. The species number densities were calculated from these signals using calibration measurements and the temperature was deduced from the total number density via the ideal gas law.
The simultaneous detection of all major species with each laser pulse also enabled the determination of the instantaneous mixture fraction [23] following the method by Bilger et al. [24].
400 single-shot Raman measurements per measurement location were performed at a total of 70 measurement locations at axial (8 mm < z < 80 mm) and radial (-3 mm < x < 27 mm) locations.
Measurement locations with x < 8 mm and z > 27 mm were not accessible due to clipping of the solid angle of the detection optics.
Systematic uncertainties were typically ±3 -4% for the temperature and mixture fraction, ±3 -5% for the mole fractions of O2, H2O, CO2 and CH4. Because of the low concentrations of H2 and CO in the flames investigated, the uncertainty is relatively large for these species. These are estimated to be around ±20% at a mole fraction of 0.01. The statistical uncertainties were approximately 30% larger than stated in a previous study [14] due to a lower pulse energy applied in the current measurements.
Typical statistical uncertainties (for a single shot measurement) were 3% for the temperature and mixture fraction, 4% for H2O and 9% for O2 and CO2 in the exhaust gas [25,14].

Acoustic Measurements
Pressure oscillations in the combustion chamber and in the two air plenums were measured using calibrated microphone probes (Brüel & Kjaer, type 4939), with a sampling rate of 100 kHz. The microphone probes were calibrated for frequencies up to 10 kHz. The pressure power spectrum at each location was computed by slicing the long-duration pressure signal into one-second segments, and calculating the power spectrum for each segment. Afterwards, the spectra of the segments were averaged, resulting in a frequency resolution of 1 Hz. Acoustic modes leading to an asymmetric pressure distribution in the combustion chamber, such as transversal modes or asymmetric flow features like precessing vortex cores (PVCs), can be detected by using the difference signal of two microphone probes mounted on opposite sides of the combustion chamber, but at the same axial position.

Characterization of the Flames
An overview of the heat release zone for the different studied stratification ratios is given in Figure 3. Here, the mean and standard deviation of the OH * chemiluminescence (CL) signal are shown. The flame shape is similar for all operating conditions. All flames feature a V-shape and the apex is approximately 90° for all flames. The flame root is lifted approximately 10 mm from the nozzle, while the main flame zone is between 30 mm and 40 mm above the nozzle. The standard deviation for the OH * CL however differs for the three studied operating conditions. While for S = 1.0, the standard deviation seems more homogenously distributed, for S = 1.38 and S = 1.67, an area with low OH * CL fluctuations is observed between the inner shear layer (ISL) and inner recirculation zone (IRZ), which becomes more pronounced with increasing S. Apart from this feature, however, all studied flames behave similarly. The origin of those fluctuations will be discussed below.

Thermo-Acoustic Oscillations
All examined flames feature strong self-excited thermo-acoustic oscillations. The overall sound pressure level in the combustion chamber increases with increasing stratification ratio from p' ≈ 370 Pa at S = 1 to p' ≈ 460 Pa at S = 1.67. Several distinct modes with different amplitudes occur in the frequency spectrum of both the pressure in the combustion chamber and two plenums as well as in the integrated OH * CL signal. Table 2 shows an overview of all observed thermo-acoustic modes at different locations within the combustor and for different measurement quantities, as well as the physical origin of the individual modes. Furthermore, the trend of the individual modes with increasing S is depicted.
The identification of the individual modes is based on, numerical modelling of the Eigenmodes of the combustion chamber and plenums, which was performed in a previous study [15] in order to identify the origin of different acoustic Eigenmodes of the system. HOP corresponds to the Helmholtz-mode of the outer plenum with the combustion chamber, OP1, OP2, IP1, and IP2 are Eigenmodes of the inner and outer plenum, respectively. PVC is the mode corresponding to the precessing vortex core. The frequency spectra of the pressure oscillations in the combustion chamber as well as in the inner and outer plenum for the perfectly premixed flame with a stratification ratio S = 1 are shown in Figure 6. The frequency spectra of the acoustic oscillations in the combustion chamber for the different stratification ratios are shown in Figure 6b. Here, the spectra are similar in shape. However, the amplitudes at some characteristic frequencies differ for the different stratification ratios, mainly at 180 Hz.
A good overview of the (global and local) dynamic behavior of the different flames is gained by performing a pixel-wise fast Fourier transform (FFT) of the OH * CL signal, which is shown in Figure   7.

Thermo-Acoustic Feedback Regions
Next, the distinct modes of the flame shape, represented by the OH LIF distributions, are considered.
To give an overall impression of the shape of the OH distribution, Figure 10    It was clearly seen that the stratification influenced the level and the spectral composition of the sound emissions. The dominant mode in the outer shear layer and outer recirculation zone is at 700 Hz, i.e. the strongest thermo-acoustic mode. It is assumed that this mode exhibits the highest fluctuation amplitude, because it is coupled to the outer plenum from where most of the fuel is supplied. Also, the decrease of OH * signal of this mode with increasing S corresponds to the decrease of fuel in the outer plenum. In future work, it will be investigated how the spectral distribution and the attenuation of acoustic emissions can be specifically controlled by stratification.

Acknowledgement
The financial support by the German Research Council (DFG) within the Collaborative Research Center 606 is gratefully acknowledged. Michael Severin and Yawei Gao are acknowledged for their help in setting up the experiment.