Fuel variability effect on flickering frequency of diffusion flames

Abstract

It is known that fuel variability of different gas suppliers may cause combustion instability in a gas turbine combustor. Mechanisms that control the time scale of the heat release oscillations and acoustic pressure perturbations are both physical and chemical in nature, and thus can be influenced by changes in fuel composition. The intent of this study is to investigate the fuel variability on the flickering frequency of diffusion flames in the hope of understanding some of the fundamental aspects of fuel variability effect on the dynamics of combustion. Experiments were conducted at atmospheric pressure with a matrix of methane and propane blends. An optical fibre system was applied to capture simultaneously the flame flickering at two different light frequencies (430 nm and 516 nm), which provided a means of comparing the chemistry change. It was found that the low frequency oscillation of flow and flame structures depended only weakly on the exit velocities of the fuel, while ambient conditions had a significant effect on flickering frequencies and spectrum. The results of using CH₄ and C₃H₈ as test fuels at different flow rates showed very little variations, with peak frequencies at 11-13 Hz. When the jet flame was not disturbed, harmonics to at least the third mode were obtained in most of these cases. However, the cases which included CH₄/C₃H₈ splits of 90/10, 85/15 and 80/20 by volume showed that unstable flickering frequencies and flame harmonics were not observed. When a mixture of methane/propane at a ratio of 1:1 was used the peak flickering frequency was around 6 Hz, and slight disturbance in the environment would cause the harmonics to disappear. Mechanisms thought to produce changes in the dynamic response and frequency harmonics were discussed.

Keywords

fuel variability / flickering frequency / diffusion flames / dynamics of combustion

Introduction

Large-scale research efforts were focused on instabilities of diffusion flames to determine the fundamental mechanism of flame oscillations and their effects on mixing. It was known that buoyancy would cause flame front to regularly oscillate at a low frequency, typically ranging from 10 to 20 Hz [1]. In numerical studies of flickering flames, it was shown that the shear layer of burnt gases had an inflexional velocity profile and was therefore unstable in the sense of Kelvin-Helmholtz instability [2-6]. Wang et al. [7] addressed the important issue of resonance by conducting a systematic analysis of forced oscillation in planar diffusion flames under weak external force. According to their study, diffusive thermal resonance occurred between the imposed flow oscillations and the intrinsic flame oscillations.

The evolution of vertical flow structures in buoyant jet diffusion flames, their interaction with flame structures and the low frequency flame oscillations were studied extensively [3,5,6,8-13]. It was concluded that flickering motion formed via coupling of the fluttering with the movement of vortex rings. The axisymmetric, low frequency oscillation of flow and flame structures only depended weakly on the type of fuel, the fuel nozzle size, and the exit velocities of the fuel and the air coflow [.,6,8]. Furthermore, vortex interactions with flames formed the basis for understanding naturally flickering buoyancy-dominated diffusion flames [14]. However, the fuels chosen were predominantly single composition and there was a lack of investigation on fuel variability effect, especially on fuels with mixed composition, which might have very different impact on important parameters such as Lewis number.

Durox et al. [9] used two types of burners to determine the flickering frequency. A photomultiplier was adjusted to aim at a point on the screen equivalent to the area over which the flame front passed during its oscillatory motion. The experimental results were in agreement with the works of Lingens et al. [5,6] who noted only a weak influence of the fuel velocity upon the flickering frequency of diffusion flames at small jet velocities. The experimental works of Cetegen and Dong [11] on large scale dynamic behaviour diffusion flames found that buoyant diffusion flames originating from circular nozzles exhibited two different modes (sinuous and varicose) of flame instabilities. The experiments pointed towards the feasibility of altering buoyant flame behaviour under the earth's gravity. Experiments on predicting buoyancy-induced flame flickering frequency were conducted under the swirling flow conditions produced by a rotating Bunsen burner [15,16]. Four discrete peaks were reported on the frequency spectrum for the flame flickering with v=0.6 m/s and n=1200 r/min. The observed harmonic frequencies indicated that the flame flickering frequency had a clearly defined periodicity [15].

Flame stability under various gravity fields was also investigated to clarify the buoyancy effects on combustion phenomena [13,17,18]. Katta et al. [13] investigated the growth behaviour of two instability waves which consisted of inner and outer vortices in a jet diffusion flame. They showed that the flickering frequency decreased in weak gravity fields, which was due to the reduction of the velocity gradient in the shear layer of the jet. The experiments conducted by Sato et al. [17,18] were under wide conditions regarding the Reynolds number and gravity levels. It was found that there were two different flickering modes (tip flickering and bulk flickering), which were characterized by the Froude number coupling with the buoyant force. They suggested that flickering phenomena due to forced convection, for which frequency was affected by the velocity, would be observed at relatively low velocity conditions.

Fuel quality and composition impact turbine life and availability as well as combustion instability. Chemical and physical characteristics of fuel determine combustion characteristics, and thus the performance of gas turbine combustor is influenced by changes in fuel composition. Changes in combustion instabilities were observed in recent studies, which investigated the sensitivity of dynamically unstable test rigs to changes in fuel composition and heat content for fuels containing high concentrations of propane [19,20]. However, only few quantitative investigations were conducted on the fuel variability effect on flickering frequency of jet diffusion flames as pointed out above.

Since efficient measurement techniques that are able to display peak frequency and flickering frequency spectrum on site were not available in the old days, there was a lack of study on fuel variability effect on the flame flickering frequency. Therefore, it seems that there is a need to investigate further the flickering phenomenon of a jet diffusion flame using automated data collection and analysis system. The intent of this study, then, is to investigate the fuel variability on the flickering frequency of diffusion flames in the hope of understanding some of the fundamental aspects of fuel variability effect on the dynamics of combustion.

Experimental setup

The overview of the experimental set-up of the test rig is shown schematically in Fig. 1. It consists of a burner system, a data collection system and a bifurcated fibre optic system for the sensing of light emissions at two chosen wavelengths.

The burner, vertically supported by a traverse gear for easy positioning, had a nozzle of 5 mm inner diameter. The burner had been applied for impinging flame investigations whose details could be found in Refs. [21,22]. Fuel pipes connected the burner with two compressed gas cylinders, which contained methane and propane respectively. Each fuel was regulated by a control valve and measured by a calibrated rotameter. A bifurcated fibre optic system which consisted of a focusing lens, fibre optic cables, monochromatic filters, and photomultipliers was used to simultaneously measure the flame light emissions at two chosen wavelengths. Two equal channels of light signals of the same intensity from the same image volume were guided to two photomultipliers (ORIEL model 70704). Two chosen filters were positioned at the end of each channel. The CH* and C₂* interference filters at wavelengths of 430±5 nm and 516±2.5 nm were used respectively. Obviously, what was measured were the summation of the soot light and chemiluminescence emissions of CH* and C₂* at the two chosen wavelengths for a diffusion jet flame. Nevertheless the availability of the two wavelengths may provide qualitative information on flame chemistry. As shown later in the paper, the system is able to pick up the significant change in flame chemistry. National Instrument DAQ card (model PCI-MIO-16E-1) and LabVIEW 8.2 software was applied for data acquisition, monitoring and analyses of the flickering diffusion flame. This measurement system enables the fast and effective measurement and the online observation of the flame flickering frequency spectrum. As a result, very effective experimentation and tests can be conducted.

Methane and propane were used as fuel. The flow rates were controlled by rotameters whose accuracy was 2% of the full scale. The flow rates were set between 1.67×10^-5 m³/s and 5×10^-5 m³/s for the methane and between 0.36×10^-5 m³/s and 3.61×10^-5 m³/s for the propane, these values being given at standard conditions. Test cases which used CH₄, C₃H₈, CH₄/C₃H₈ mixtures at various flow rates were conducted under atmospheric pressure. The data could then be analyzed by LabVIEW software to obtain important information such as the peak flickering frequency and spectrum.

Results and discussion

Figure 2 shows the typical jet diffusion flames taken at various camera shutter speeds. The top two rows are sequences of images taken at five frames per second (fps) and the shutter speed is set at 1/1000 s. In Fig. 2(a), the fuel is a mixture of propane and methane with a mixing ratio of 1∶1. The fuel flow rate is 2 L/min. Flame necking is clearly seen. The third row is a sequence of images of the same flame jet taken at 500 fps and at a 1/500 s shutter speed. Obviously, more detailed development of the flame dynamics is shown. The bottom row is a sequence of images of methane jet flame taken at 500 fps with the same flow rate as the middle row case. It can be seen that the flame is much thinner. This is doubly confirmed by repeated imaging and observation. Again flame necking is observed. It is worth mentioning that these necking processes cannot be observed by direct visualization because the time scale is too small for human eyes to resolve. Figure 2(c) is an image taken at 1/10 of a second, which is about the same perception as human eyes.

In Fig. 3, sequences of diffusion flame images taken at a range of camera shutter speeds (1/2000 s to 1/10 s) are illustrated. The test fuel is propane at a constant jet velocity of 2 L/min. The flame necking development can be clearly seen from Figs. 3(a), (b) and (c). As the shutter speed decreases, vortex structure and flame necking processes are becoming less obvious. Again, the last row of images (Fig. 3(h)) is taken at a shutter speed of 1/10 s, which shows the visualization of the flames with human eyes.

For more quantitative measurements, attention is paid to the statistical measurement of the flickering frequencies and their corresponding flame flickering spectra. Although naturally occurring flickering flames are difficult to investigate experimentally due to cycle-to-cycle variations, the availability of online display of the flame flickering spectrum makes it possible to conduct very effective testing and observations. The peak frequency can be determined online by the built-in algorithm of the LabVIEW software. It has been observed that the flame flickering frequency is very sensitive to the minute changes in ambient conditions such as heavy footsteps and loud conversations in the laboratory or next door. On the other hand, the flickering frequency is little affected by the change in fuel jet velocity. As shown in Table 1, the peak frequencies obtained from each set of test cases have little variations, even though the fuel flow rates changed three times. These data were taken when the disturbance to the laboratory was at a minimum. One possible reason to explain the strong dependence on ambient condition but not on the jet fuel velocity is that velocity-induced vortex/vortices only require a limited amount of velocity gradient from the fuel jet side and extra fuel velocity gradient will not necessary impart more energy to the vortices which play crucial roles in bringing fresh air to mix with the fuel. A local vortex may affect the local burning rate significantly due to the ability of bringing fresh air to the flame front, which could be the most important mechanism of flame necking in a diffusion flame jet. When a disturbance is generated in the surroundings, the air may be entrained in such a way that the vortex ring may be broken up. As a result, the local air supply will be affected, which causes significant change in flame shape and dynamics.

Phase diagrams of the two signals at the chosen C₂* and CH* wavelengths are illustrated in Fig. 4. The signal collected (S_t) at each channel is the summation of chemiluminescence emission (S_c) and soot light emission at that wavelength (S_s). Therefore, there is S_t= S_c+S_s. If S_c <<S_s, the soot signal would play a dominant role in deciding the flame flickering spectrum, which means that the ratio change of the CH^* and C₂^* in the reaction zone will not have much impact on the two signals. It is expected that the phase diagram of the two signals would be in phase. Figure 4 clearly shows that the soot signal is not dominant in the pure methane case. The global signal at CH^* wavelength is stronger than that at the C₂^* wavelength statistically. Interestingly, both signals give very similar flame flickering spectra.

Flame flickering harmonic frequencies have been observed in Figs. 5(a), (b) and (c). As much as six harmonics are observed (see Fig. 4(a)). There is a lack of in-depth investigation on the formation mechanism of the harmonics. It may be caused by the coupling of the external force (buoyancy forced fluctuation) with the intrinsic flame oscillations [7].

Figure 5(c) shows that the peak frequency is halved compared with those in Fig. 5(a) and (b) though the fuel jet velocity is almost the same as those in Figs. 5(a) and (b). This observation is in contradictory to the theoretical work reviewed in Ref. [9]. It is believed that the natural oscillation of buoyant diffusion jet flames is induced by the unstable motion of the outer large structures surrounding the flame front which is the consequence of hydrodynamic shear instability driven by convective effects. As a result, the chemical reaction and Lewis number should have little effect on this type of instability. The experimental evidence presented in this paper seems to indicate that the phenomena are more complex in nature, especially for mixed fuels. The hydrodynamic induced instability alone may not explain this significant change. The different chemical compositions are more likely to be the main cause for this dramatic change. One possibility is that the different Lewis number of the two fuels has induced diffusive-thermal instability and this is coupled with the buoyancy induced instability. It is also possible that the fuel mixture has very different chemical reaction. The high speed images in Fig. 2 also indicate that flame structure and dynamics are very different for the methane and mixed fuel jet flames. Further in-depth study is necessary.

In most cases, the flame dynamics at the two frequencies are very similar. However, from Figs. 5(d), (e) and (f), it can be seen that the dominant flickering frequency measured at the two wavelengths are different. It has also been observed that the flickering frequencies are not locked to particular values, which indicates that the flame is not in regular oscillating mode.

Table 2 shows that the mean peak frequency of the mixed fuel jet flame is a mild function of the mean fuel volume flow rate (jet velocity), which is different from the case of methane and propane jet flame presented in Table 1. Why this is the case is worth further investigation. On the other hand, the data consistently show the approximately halved peak frequency compared with the cases of pure methane or propane. The standard deviation of the flame flickering peak frequency is small.

Conclusions

An automated fibre optic system has been applied to systematically investigate the flame flickering frequency of jet diffusion flames. The tests have shown that global emissions at selected wavelengths could capture the flame flickering frequency successfully. The flame images taken at various shutter speeds using different cameras suggest that the necking processes cannot be observed directly with human eye visualization. In most test cases, the standard deviation of the flame flickering peak frequency is small if the ambient condition is under good control. When the jet flame is not disturbed, harmonic frequencies to at least the third mode are observed in many test cases. In particular, for the methane cases, up to six harmonic frequency modes are observed. The fuel jet velocity is found to have little effect on the flickering frequency. In contrast, it has been found that the ambient condition has significant effect on the flickering frequency spectrum. It has also been observed that the fuel variability does have strong effect on the flame flickering frequency.

References

[1] Chamberlin D S, Rose A. The flicker of luminous flames. Ind Eng Chem, 1928, 20(10): 1013-1016

[2] Toong T-Y, Salant Richard F, Stopford John M, . Mechanisms of combustion instability. Symposium (International) on Combustion, 1965, 10(1): 1301-1313

[3] Buckmaster J, Peters N. The infinite candle and its stability-A paradigm for flickering diffusion flames. Symposium (International) on Combustion, 1986, 21 (1): 1829-1836

[4] Ellzey J L, Oran E S. Effects of heat release and gravity on an unsteady diffusion flame. Symposium (International) on Combustion, 1990, 23(1): 1635-1640

[5] Lingens A, Neemann K, Meyer J, . Instability of diffusion flames. Symposium (International) on Combustion, 1996, 26(1): 1053-1061

[6] Lingens A, Reeker M, Schreiber M. Instability of buoyant diffusion flames. Experiments in Fluids, 1996, 20(4): 241-248

[7] Wang H Y, Bechtold J K, Law C K. Forced oscillation in diffusion flames near diffusive-thermal resonance. International Journal of Heat and Mass Transfer, 2008, 51(3, 4): 630-639

[8] Chen L D, Seaba J P, Roquemore W M, . Buoyant diffusion flames. Symposium (International) on Combustion, 1988, 22(1): 677-684

[9] Durox D, Yuan T, Villermaux E. The effect of buoyancy on flickering in diffusion flames. Combustion Science and Technology, 1997, 124(1): 277-294

[10] Smyth K C, Shaddix C R, Everest D A. Aspects of soot dynamics as revealed by measurements of broadband fluorescence and flame luminosity in flickering diffusion flames. Combustion and Flame, 1997, 111(3): 185-194

[11] Cetegen B M, Dong Y. Experiments on the instability modes of buoyant diffusion flames and effects of ambient atmosphere on the instabilities. Experiments in Fluids, 2000, 28(6): 546-558

[12] Lee K M, Kim T K, Kim W J, . A visual study on flame behaviour in tone-excited non-premixed jet flames. Fuel, 2002, 81(17): 2249-2255

[13] Katta V R, Roquemore W M. Role of inner and outer structures in transitional jet diffusion flame. Combustion and Flame, 1993, 92(3): 274-278

[14] Papadopoulos G, Bryant R, Pitts W. Flow characterization of flickering methane/air diffusion flames using particle image velocimetry. Experiments in Fluids, 2002, 33(3): 472-481

[15] Gotoda H, Ueda T, Shepherd I G, . Flame flickering frequency on a rotating Bunsen burner. Chemical Engineering Science, 2006, 62(6): 1753-1759

[16] Gotoda H, Maeda K, Ueda T, . Periodic motion of a Bunsen flame tip with burner rotation. Combustion and Flame, 2003, 134(1,2): 67-79

[17] Sato H, Amagai K, Arai M. Diffusion flames and their flickering motions related with Froude numbers under various gravity levels. Combustion and Flame, 2000, 123(1,2): 107-118

[18] Sato H, Amagai K, Arai M. Flickering frequencies of diffusion flames observed under various gravity fields. Symposium (International) on Combustion, 2000, 28(2): 1981-1987

[19] Ferguson D H, Straub D L, Richards G A, . Impact of fuel interchangeability on dynamic instabilities in gas turbine engines. In: 5th U.S. National Combustion Meeting, Western States Section/Combustion Institute (WSS/CI), San Diego, CA, 2007

[20] Straub D, Ferguson D, Casleton K, . Effects of propane/natural gas blended fuels on gas turbine pollutant emissions. In: 5th US Combustion Meeting, Western States Section/Combustion Institute (WSS/CI), San Diego, CA, 2007

[21] Foat T, Yap K P, Zhang Y. The visualization and mapping of turbulent premixed impinging flames. Combustion and Flame, 2001, 125(1,2): 839-851

[22] Ng W B, Zhang Y. Stereoscopic imaging and reconstruction of the 3D geometry of flame surfaces. Experiments in Fluids, 2003, 34(4): 484-493

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg