Characterization of flameless combustion of natural gas in a

Characterization of flameless combustion of natural gas in a laboratory scale furnace. S. Murer*, B. Pesenti and P. Lybaert Thermal Engineering & Combustion Laboratory, Faculté Polytechnique de Mons, Rue de l'Epargne, 56, B-7000, Mons, Belgium Abstract The purpose of this study is to analyse, both experimentally and with CFD modelling, the flameless combustion of natural gas. Tests have been performed on a 30 kW laboratory scale combustor, equipped with an electrical air preheater. The influence of excess air, furnace temperature and air preheating temperature on reaction zone volume and location has been examined through flame UV emission images and the impact of the same parameters on outlet NOx emission is also observed. The influence of combustion and turbulence models on CFD results is discussed. This work is part of a larger research program devoted to the study of flameless combustion techniques and their application to various fuel/oxydant combinations. 1. Introduction Diluted combustion, also called “Mild combustion “ or “Flameless oxidation” in literature, is known since the early nineties as a very effective way to abate NOx emissions in high temperature gas fired furnaces using highly preheated combustion air (HiTAC technologies) [1,2]. Besides NOx emissions, the reaction zone and heat transfer characteristics in the furnace are also deeply modified by this particular combustion mode [3, 4]. Therefore fundamental studies are still needed to better understand the reaction process in diluted mode and develop models able to predict furnace performance. In the present work, the flameless combustion of natural gas with highly preheated air is studied in a laboratory scale furnace (30kW), experimentally and with CFD techniques. 2. Experiment 2.1. Experimental setup Figure 1 shows the experimental high temperature air combustion setup used in this study. The test rig is made up of a vertical combustion chamber of square cross section. The walls are lined by a rigid fibrous insulation layer, 100 mm thick on the lateral walls and 200 mm thick on the upper and lower walls. The inner dimensions of the chamber are 0.35 x 0.35 x 1 m³ (width x depth x height). The combustion air is preheated in an electrical heater, its temperature can be adjusted up to 1000°C. It is then introduced axially at the base of the chamber through a 25 mm diameter nozzle. Natural gas is injected through two offaxis injectors, 3 mm diameter, placed at 160 mm on both sides of the air nozzle and inclined at about 11° with respect to the furnace axis (figure 2). The flue gases are extracted at the top of the chamber through 12 circular lateral openings and are cooled in a collector before being released through the chimney. ____________________________ *Corresponding author: sebastien.murer@fpms.ac.be Associated Web site: http://stecwww.fpms.ac.be Proceedings of the European combustion Meeting 2005 Water cooled tubes Flue gases collector Combustion chamber Air preheater Figure 1: Experimental setup. The combustion chamber temperature is controlled by 4 vertical water cooled bayonet tubes, placed in the corners of the chamber. By adjusting the immersion depth of the tubes, the furnace temperature can be made to vary continuously from 900°C up to about 1300°C. The setup is equipped to record the following variables: - Air and fuel flow rates are measured by orifice meters. - Preheated air, fuel gas and cooling water temperatures are measured by thermocouples. - The wall temperature profile along the height of the furnace is measured by 8 thermocouples mounted flush to the inner surface of the lining. The first thermocouple is placed at 85 mm from the base, and the other ones are distant by 90 mm from each other. (figure 3). - Flue gases temperature at furnace outlet is measured by a suction thermocouple probe located in front of an outlet opening. - The flue gases at furnace outlet are also sampled by a cooled probe and their composition is measured by a set of analysers. The following species concentrations (dry basis) are measured: O2 by a paramagnetic analyser, NO/NOx by a chemiluminescence analyser, CH4, CO, and CO2 by IR analysers. Moreover, the combustion chamber is equipped with a 50 mm x 640 mm (width x height) vertical quartz window (figure 3) in order to perform flame imaging. A LaVision intensified CCD camera equipped with a narrow band pass filter centred at 307 nm was then used to record UV self emission of OH radicals which are considered as good markers to observe the shape and location of the reaction zone [5]. 2.2. Experimental conditions Four series of tests have been performed, with the following operating conditions: fuel heat release Φg = 30 kW, preheated air temperature Ta = 800°C and 1000°C, excess air ratio E = 10% and 20%. Natural gas was used, with the following composition (volume basis): CH4 = 90.85%, C2H6 = 4.80%, C3H8 = 0.97%, C4H10 = 0.41%, CO2 = 0.79% and N2 = 2.02%. During each series of tests, the furnace temperature was made to vary by changing the immersion depth of the cooling tubes. The corresponding furnace temperatures Tf measured at furnace outlet, ranged from 1050°C to 1300°C. 3. Experimental results Figure 2: Sketch of the injectors. Flue gases collector Twall T8 Quartz window T1 Figure 3: Combustion chamber. 3.1. Flame shape and OH emission profiles No visible flame was observed during the tests. On the other hand, instantaneous UV images show a fluctuating combustion zone, at some distance from the bottom of the furnace. Each recorded picture is an average of 50 instantaneous images. Figure 4 shows typical average UV emission images obtained for a fuel heat release of 30 kW, air preheating temperature of 800°C and furnace temperature around 1050°C. Figure 4-A shows the picture obtained for 20% excess air. In this case, two different zones with maximum OH emissions are observed. They are more or less elliptical in shape and are located symmetrically at some distance of furnace axis, where the air and gas jets are expected to meet, giving a reaction zone of symmetrical shape, detached from air nozzle exit. Figure 4-B gives the picture obtained for 10% excess air. Compared with the former case (E = 20 %), a single combustion zone with lower emission levels is obtained. The volume and detachment distance of the combustion zone are also increased. A quantitative comparison of the effects of air preheating temperature and excess air is given in figure 5, which shows the emission profiles obtained for a furnace temperature of 1100°C. In each case, the profile has been taken along the vertical line passing through the point of maximum intensity. 2 The vertical distance zmax at which maximum emission is measured is related to the location of the reaction zone. This distance is plotted as a function of furnace temperature Tf in figure 6. The figure shows that the distance decreases when furnace temperature increases. Above Tf=1070°C for E=20% and Tf=1140°C for E=10%, the distance becomes constant (about 32 cm), independent of furnace temperature, excess air, and air preheating temperature. This minimum distance should be related to the particular geometry of the jets. The combined effect of furnace temperature and excess air on zmax could then be explained by the relative influences of turbulent diffusion and reaction kinetics on reaction zone volume and location. 80 70 (A) E=20 %, Tf=1050°C (B) E=10%, Tf=1100°C Zmax [cm] 60 50 40 30 20 10 0 1050 1100 1150 Figure 4: UV imaging (50 pictures average) Φg = 30 kW, Ta = 800°C. Figure 5 shows that excess air has a significant effect on the combustion zone location, shape and emission level: a higher excess air leads to a shorter distance zmax between maximum emission point and furnace bottom, and higher UV emission levels. A higher air preheating temperature has a similar effect, which is more pronounced at low excess air. E=10% - Ta= 800°C E=10% - Ta=1000°C E=20% - Ta= 800°C E=20% - Ta=1000°C Tf (°C) 1200 1250 1300 Figure 6: Location of maximum emission. 500 450 400 350 300 Counts 250 200 150 100 50 0 0 10 20 30 40 50 60 70 80 E=10% - Ta=800°C E=20% - Ta=1000°C E=20% - Ta=800°C E=10% - Ta=1000°C Axial distance (Z) [cm] Figure 5: OH emission profiles for Tf = 1100°C. 3 [NO] dry basis @ 3%O2 (ppm) 3.2. Wall temperatures The normalized wall temperature profiles are plotted in figure 7 for a furnace temperature of about 1100°C. For E=20%, they show a distinct local maximum measured at the third thermocouple location T3 (i.e. for z=265 mm) and a local minimum at T4 (z=355 mm) for Ta=1000°C and T6 (z=535 mm) for Ta=800°C. For E=10%, this maximum almost completely disappears and the wall temperature continuously increases from T1 to T8. These profiles can be related to the UV emission profiles as both kinds of profiles depend on the reaction zone volume and location. They confirm the dominant effect of excess air on these characteristics. Measured NOx emissions are very low and increase with furnace temperature (figure 9), ranging from a few ppm for Tf=1050°C up to about 30 ppm for Tf=1300°C. For given excess air and air temperature, the relationship between NOx concentration and furnace temperature is linear. For a given furnace temperature, NOx emissions increase with excess air, with air temperature having but a small effect. 30 25 20 15 10 5 0 1050 E=10% - Ta= 800°C E=10% - Ta=1000°C E=20% - Ta= 800°C E=20% - Ta=1000°C (Tw-Twmin)/(Tf-Twmin) 0,45 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 0 E=10% - Ta=800°C E=10% - Ta=1000°C E=20% - Ta=800°C E=20% - Ta=1000°C 1100 1150 1200 1250 1300 Tf (°C) Figure 9: NO (dry basis @ 3%O2) concentration at furnace outlet 20 40 60 80 Axial distance [cm] 4. Numerical simulation Simultaneously, CFD simulations of the combustion chamber have been performed using FLUENT 6.1. 4.1. Mesh Thanks to symmetry, only a quarter of the furnace was modelled. The 12 outlet openings were replaced by a single slit having the same cross section. In the first simulations presented hereafter, the immersion depth of the water cooled bayonet tubes was set equal to 500 mm. Figure 10 shows the unstructured mesh used in the cross section of the furnace. The 3-D mesh was obtained by extruding this 2-D mesh in the z-direction. The created mesh has 79769 elements. Figure 7: Normalized wall temperature profiles –Tf = 1100°C . 3.3. NOx and CO emissions Measured CO content of flue gases at furnace outlet is shown in figure 8 as a function of furnace temperature. For E=20%, it is about 30 ppm (dry basis) and is almost independent of furnace temperature. For E=10%, CO content increases with the temperature, from about 30 ppm at 1050°C up to about 80 ppm at 1300°C. Air preheating temperature has a negligible effect on CO emissions. 90 80 70 E=10% - Ta= 800°C E=10% - Ta=1000°C E=20% - Ta= 800°C E=20% - Ta=1000°C [CO] (ppm) 60 50 40 30 20 10 0 1050 1100 1150 1200 1250 1300 Figure 10: Cross section of the mesh. 4.2. Reference case Up to now, simulation results have been obtained for the following operating conditions: Φg = 30kW, Ta = 1000°C, E = 20% and immersion depth equal to 50 cm. Tf (°C) Figure 8: CO concentration at furnace outlet. 4 For the reference solution, the following models have been used [6]: - Turbulence has been modelled using the Launder and Spalding's standard k-ε model with standard wall functions. - Combustion has been modelled by the “EDM + Finite Rate” model. In this model, the reaction rate is computed as the minimum value of a mixing rate governed by the large eddy mixing time scale and the Arrhenius rate, governed by chemical kinetics. The chemical reaction is simulated by a two-step reaction scheme with CO as intermediate species. The results of these models are given in figure 11 to 13. The temperature distribution (figure 11) shows that the temperature is fairly uniform inside the furnace, with a maximum value of 1337°C obtained at about one third of furnace height. Combining this temperature distribution with CO mole fraction field (figure 12) allows to observe that the combustion takes place where the fuel mixes with the air coming from the air injector, i.e. in the vicinity of furnace axis at a distance of about 300 mm of the bottom, but also at some distance from the axis and in the lower part of the furnace, where the fuel jet meets the low oxygen content recirculated flue gases. Figure 11 and 12 also show that no reaction takes place along furnace axis and that the zone of most intense reaction is located symmetrically on both sides of the axis. Both the shape of the combustion zone and its location are compatible with the experimental UV imaging results. Temperature uniformity and reaction zone shape and location are strongly related to flue gas recirculation. Figure 13 shows the shape of the recirculation zone, i.e. the zone in which negative axial velocities are obtained. Large recirculation velocities are obtained, up to about 8 m/s for an air velocity of about 80 m/s at nozzle exit. The computed velocity field also allows to compute the recirculation ratio, which is given by the ratio of recirculated mass flow rate to the inlet air and fuel mass flow rate. This ratio varies along furnace height; its maximum value is about 5.5 at mid-height of the furnace (figure 14). Figure 13: Recirculation zone (negative z-velocities). (standard k-ε / EDM + Finite Rate) 7 Recirculation ratio 6 5 4 3 2 1 0 0 20 40 realizable k-e RSM standard k-e 60 80 100 Axial distance [cm] Figure14: Axial profile of recirculation ratio. Figure 11: Temperature distribution (K). (standard k-ε / EDM + Finite Rate) 4.3. Effect of the turbulence model Two other turbulence models implemented in FLUENT have been tested: the realizable k-ε model and the Reynolds Stress Model (RSM). The results are qualitatively very similar to the reference case. Table 1 gives some quantitative results, i.e. the maximum values of temperature, CO content and recirculation velocity. Turbulence Model Standard k-ε Realizable k-ε RSM Figure 12: CO mole fraction distribution. (standard k-ε / EDM + Finite Rate) T max (°C) 1337 1359 1364 CO max (ppm) 4170 5430 6770 Uz max (U<0, m/s) -7,24 -7,42 -6,77 Table 1 – Effect of turbulence model 5 The main impact of turbulence model is on the axial recirculation ratio profile, which is represented in figure 14 for the three turbulence models. The lowest recirculation rates are obtained with the RSM model (Rmax = 4.6) and the highest values are given by the realizable k-ε model (Rmax = 6). 4.4. Effect of the combustion model Two other combustion models have been combined with the standard k-ε model: the Eddy-Dissipation Model (EDM), which doesn't take into account any chemical kinetics limitation and the PDF model, based on the computation of a transport equation for the mixture fraction and its variance; the gases temperature and composition are related to the mixture fraction values through chemical equilibrium [6]. Combustion Model EDM+finite rate EDM PDF T max (°C) 1337 1334 1297 CO max (ppm) 4170 5740 108000 Uz max (U<0, m/s) -7,24 -7,24 -7,15 air and gas jets are expected to meet, giving a reaction zone of symmetrical shape, detached from air nozzle exit. Both reaction zone volume and detachment distance increase when furnace temperature decreases. Measured NOx concentrations are very low, ranging from a few ppm for Tf = 1050°C up to about 30 ppm for Tf = 1300°C. Excess air ratio and preheated air temperature have but a small effect on NOx concentrations. In the numerical study, 3 turbulence models and 3 combustion models have been compared. The EDM and EDM+Finite Rate models, combined with the standard k-ε turbulence model, lead to qualitatively correct results. Change of turbulence model does not affect the results very much, while using a PDF combustion model leads to large and probably unrealistic local CO concentrations. Both EDM models compute plausible CO and temperature values, with small differences in the near injection zone when kinetic limitations are considered. Local velocity, temperature and concentrations measurements will be performed in the future to validate the numerical approach. Acknowledgments This work has been performed in the framework of a research program funded by the Walloon Government. The authors wish to thank the Walloon regional authorities for their financial support. Nomenclature E Excess air ratio R Recirculation ratio T Temperature (°C) U Velocity (m/s) Φ Heat rate (kW) z Axial distance (cm) Subscripts a air f furnace, flue gases g gas w wall References [1] Hasegawa T. and Tanaka R. (1998), JSME International journal, Series B, 40 (4), 1079-1084. [2] Wünning, J.A. and Wünning, J.G (1997), Progress in energy and combustion science, 23 (12), 81-94. [3] Cavaliere A. and De Joannon M. (2004), Progress in energy and combustion science, 30, 329-366. [4] Orsino S., Weber R. and Bollettini U. (2001), Combustion Science and Technology, 170, 1-34. [5] Maurey C. (2001), Etude expérimentale de la stabilisation et du soufflage des flammes de diffusion turbulentes suspendues, PhD thesis, Faculté des Sciences et Techniques de l'Université de Rouen. [6] User's Guide, FLUENT 6.1 Documentation. Table 2 – Effect of combustion model Figure 15: CO mole fraction distribution. (standard k-ε / EDM) The combustion models are compared in table 2. The PDF model gives unrealistic CO concentrations, higher than 10 %. The solution predicted by the EDM model is not very different from the reference one. The CO mole fraction distribution obtained with EDM is given in figure 15. The largest differences are observed in the vicinity of the gas injectors. Without any chemical kinetics limitation, combustion starts at fuel inlet and this leads to a combustion zone which is attached to the fuel injectors. 5. Conclusion In this study, we experimentally investigated flameless combustion characteristics with high temperature air combustion in a laboratory scale combustor. No visible flame was observed during the tests; UV flame images showed maximum OH emissions at some distance of furnace axis, where the 6