Study of injection typology on turbulent homogeneous mixing in a natural gas swirl burner

Transcription

1 Study of injection typology on turbulent homogeneous mixing in a natural gas swirl burner Giulio Solero, Aldo Coghe, Carlo Terragni Dipartimento di Energetica - Politecnico di Milano, Italy Abstract The three dimensional gas velocity and fuel concentration fields in the turbulent mixing zone of a natural gas swirl burner have been investigated in the isothermal case by means of laser Doppler anemometry (LDA), particle image velocimetry (PIV) and laser sheet visualisation (LSV). The different techniques enabled useful insight into the fully turbulent, three dimensional swirling flow to be gained for various operating conditions. The effect of air swirl motion and fuel-air momentum ratio has been quantified on both the flowfield structure and fuel concentration distribution. The results indicate that the unmixedness index, U, may be a significant parameter to judge the mixing efficiency due to swirl strength and fuel-air momentum ratio in a specific burner geometry, at least under isothermal conditions in the near field zone downstream the quarl exit. Introduction Non-premixed swirling flows are widely used in industrial combustion systems, notably, gas turbines, boilers and furnaces, because of safety and stability reasons. Swirl increases fuel-air mixing, improves flame stabilisation and has a strong influence on flame characteristics and pollutant emissions. Although the swirling flows have been extensively used in combustion design and it is known [1] that the swirl can significantly affect the NOx emission, the relation between the combustion characteristics of swirling flames and pollutant formation still needs to be established []. Among the methodologies under development to minimise the environmental impact of this type of combustion systems, the most promising are those based on the improvement and optimisation of the mixing process between the reactants and cold fresh combustible mixture with hot gas products. It is well known [3,,5] that, independently from the combustion technology used, any improvement of combustion performance relative to pollutant formation, stability and overall efficiency requires a careful study of the mechanism of mixing and entrainment in high turbulent reacting flows. In many combustion devices both the reactants are in the gas phase and for technological reasons the coaxial geometry is commonly used to merge the two streams; the swirl motion of the main flow is used to improve flame stability and enhance mixing process [,7]. The fine structure of the resulting mixture and the mechanism of flame stabilisation, related to the recirculation zone induced by the swirling flow pattern, control the combustion process and pollutant formation. At moderate swirl intensity, the interpenetration process between the two reactants is due to a strong shear that determines the entrainment rate of the slow stream by the fast one and the formation of local stoichiometric conditions. A common feature at high swirl intensity (swirl number, S >.) is a recirculation bubble in the vicinity of the fuel jet outlet. The recirculation regime sustains the entrainment process of the outer stream into the inner one and may induce a low-frequency, precessing mode that is distinct from the jetpreferred mode. The recirculating regime also presents the capability of an efficient mixing between the streams in the region near the fuel outlet, therefore leading to a rapid 1

2 homogenisation of the combustible mixture and a shortening of the combustion chamber. The scaling of pollutant emissions in industrial flames is very difficult, because of the complex geometry of the burner and the many parameters involved. The experimental evidence is that each burner is a unique device and even small geometrical changes can influence the level of emissions. It is also known that swirling flows require detailed measurements of the flow structure and specification of the inlet conditions []. The fluid dynamic analysis is very useful to provide preliminary information on the mixing process, even if the investigation is performed under isothermal conditions. In fact, a possible useful approach is to begin with detailed measurements of the flow structure and specification of the inlet conditions for the isothermal flow, without the complications of chemical reactions, considering that the chemical kinetic time scales are usually shorter than the turbulent time scales. It is also important to know at what distance from the injector outlet the fuel has been completely mixed and diluted in the swirling air stream at the molecular level. More generally, one might be interested in characterising the turbulent flow field and the fuel concentration distribution at several cross-sections downstream of the injector outlet. The primary interest is in the near field region since this area is the location in which most of the mixing and reactions take place. Owing to the high complexity of the flow field, the experimental characterisation has been usually performed through non-intrusive optical techniques capable of quantitative point measurements and qualitative visualisations [9,1]. In the past years, we have investigated different fuel injection typologies: transverse radial and coaxial injection into a co-flowing air stream with variable swirl strength and co-flow to fuel momentum ratio [11]. This paper reports on the main results of the experimental characterisation of the isothermal mixing process in the near field of a natural gas burner; the study has been limited to this primary mixing region because the main characteristics remain almost the same in the presence and in the absence of chemical reactions, so that an isothermal analysis can provide preliminary information about the nature of the subsequent combustion process. The main parameters investigated were: air swirl and fuel to air velocity ratio. The optical techniques used were all based upon the Mie scattering principle: laser Doppler anemometry (LDA), particle image velocimetry (PIV) and laser sheet visualisation (LSV) integrated by quantitative image analysis. The ultimate goal of the research is to provide a data base to test submodels for CFD computations and to acquire an understanding of the three dimensional velocity field and fuel concentration distribution in the primary mixing zone, produced by both, flue gas recirculation and swirl, in a turbulent non-premixed swirling reactor similar to many industrial combustion systems. Experimental set-up The experimental apparatus consists in a laboratory scale of a swirl burner mounted in a fully transparent chamber, relatively large to produce little effect upon the flow pattern near the burner exit, where the isothermal conditions could be assumed reasonably representative. A schematic diagram of the burner is shown in Fig.1. The burner has a centrally positioned fuel delivery tube coaxial with the swirling air flow, with a divergent quarl exit similar to those used in many industrial burners; the swirl motion is imparted through an axial plus tangential air entry. Air was supplied by a compressor and injected by four radial and four tangential inlets located on the wall of the burner and swirl was controlled by varying the relative

3 amounts of axial and tangential air flow. For the isothermal analysis air was also supplied through the central pipe, instead of methane. The fuel is injected axially through a single-hole nozzle ( mm inner diameter) with the tip set exactly at the divergent throat. The relevant dimensions of the burner are summarised in Table 1. Fig. 1: schematic view of the investigated burner. Burner inner diameter, D 3 Fuel tube diameter, D 1 15 Co-axial nozzle diameter, d Burner quarl exit diameter, D bq Burner quarl divergence angle [deg] 3 Test chamber 3x3x5 Table 1: relevant burner dimensions [mm]. The velocity measurements have been performed by a dual-beam, two-components LDA system operated in back-scatter with a 5 Watt Argon-ion laser (Spectra Physics, mod. ), which allowed contemporary evaluation of axial and tangential or radial and tangential velocity components. Sensitivity to the flow direction was provided by a MHz light frequency shifting through a Bragg cell. Measurement volume dimensions were around.1 mm diameter by 1. mm length. Oil droplets of nominal 1 µm size were used as tracers, by adding the particles to the surrounding air stream and/or the gaseous fuel, depending on the measurement objectives. The light scattered by the particles was focused onto preamplified photomultipliers and the Doppler signals were monitored by a 5N1 PDA Dantec processor interfaced with a PC through a DMA module. At moderate seeding density, the velocity signals presented rather high validation rates, at several khz; higher seeding density was necessary for the planar visualisations. Errors in the mean axial and swirl velocities were estimated at about %, having increased the number of samples in the more turbulent regions. Moreover, a few instantaneous -D measurements of the flow field have been performed through a PIV system (TSI model), kindly provided by Sensortech, and essentially constituted by a double Nd:Yag laser and a CCD camera (1x1 pixel resolution). Images of the "frozen" particles (the same used for LDA) flow recorded in the correspondence of the two laser pulses have been processed by cross-correlation technique, to obtain the instantaneous 3

4 flow field in planes normal to the burner axis. Laser sheet visualisations were obtained by means of a pulsed copper vapour laser (Oxford Lasers, mod. CU15A) with an average power of 15 W and capable of 1 khz pulse repetition rate. Each pulse has a duration of 5 ns and a system of cylindrical and spherical lenses and mirrors provided a light sheet about 1 mm thick which intersected the burner axis normally at different downstream distances from the nozzle (see Fig. ). The images, produced by the light scattered by the same oil particles used for LDA, were recorded at 9 by a CCD camera with the exposure time set to 1 ms; each image was thus averaged over 1 consecutive laser pulses. The images were analysed and processed through the Image Pro Plus software, in order to improve image quality and contour definition after background subtraction. Further processing was obtained by means of specially developed routines. Fig. : schematic view of the experimental set-up (LDA and LSV). Experimental results The experimental analysis was performed at different axial distances from the exit plane of the burner quarl from X/R bq =.1 to 1. The standard operating characteristics of the burner are summarised in Table. As it can be seen, different values of fuel velocity and swirl number of the air stream have been tested, all correspondent to a lean flame, in combustion case. In the Table the initial air velocity is referred to the inlet cross-section of the burner quarl, while the fuel injection velocity refers to the nozzle outlet. An important parameter is the square root of the momentum ratio, denoted by M, and defined as ( ρ f U f )/( ρa U a ), where f and a denote fuel and air exit conditions. The quantity U f is the injection velocity of the fuel at the nozzle exit and U a is the axial bulk velocity of the co-axial airflow at the upstream end of the burner quarl. In homogeneous fields as is the case in the present isothermal experiments, with air used instead of natural gas, the square root of M reduces to the velocity ratio, r = U f /U a, reported in Table. This ratio is an important parameter that controls the mixing of fuel and air in the analised geometry.

5 As previously outlined, the burner can be operated at variable swirl number S of the air stream. Swirl number, S, is defined as the ratio of the flux of angular momentum to the flux of axial momentum divided by the radius of the burner [13]. The swirl number reported in Table has been evaluated by numerical integration of the axial and circumferential velocity profiles measured by the LDA along one diameter at the burner quarl inlet plane. Swirl number equal to zero corresponds to pure axial air entry in the swirl generator; while S=1. corresponds to 5% axial plus 5% tangential air entry. Initial air velocity (m/s) 7.9 Fuel injection velocity (m/s) Fuel-air velocity ratio r Swirl number, S 1. Co-flow Re x 1 1,7 Table : operating conditions of the burner. The first objective of the experimental investigation was to develop measuring techniques and data analysis procedures allowing the evaluation of the mixing efficiency for each operating condition. The selected approach was to evaluate the entrained recirculated air flow through LDA measurements and to define, as in [1, 15], an unmixedness index, U, which can be evaluated through quantitative analysis of light distribution on the planar images: c' U = C ( 1 C) where C stands for the mean fuel concentration and c is the rms value. The value C = 1 represents the condition with fuel only, while C = the case with air only. The above definition has the advantage that: U = means full mixing (c = ; C ) U = 1 maximum variance (c = C =.5) and thus worst mixing U = presence of only air (C = ) or only fuel (C = 1) The mean concentration C may be determined from the laser sheet images by spatial averaging over a selected area. Our choice was to consider the circular area centred on the burner axis with the radial extension of the quarl exit; by comparing the results at several cross-sections downstream the exit plane, the spatial evolution of the mixing process can be analysed and a numerical index used to compare different operating conditions. In fact, the value of U is only dependent on the initial calibration of the technique, which requires to maintain constant the volume concentration of tracing particles in the seeded flow and normalise the local concentration to the initial mean concentration C measured at the nozzle exit. The second step will be to determine the detailed distribution of the three mean velocity components, with related normal and shear stresses, and try to find correlation between the flowfield structure and mixing efficiency. A preliminary analysis has been completed regarding the mean and fluctuating velocity components used to characterise the flow 5

6 structure of the burner; these results will be presented first. Flowfield structure The basic flowfield structure in the primary mixing region is presented in Fig. 3 for the swirl number S = and for different values of fuel-air velocity ratio. Fig. is referred to swirl number S = 1. at the same values of fuel-air velocity ratio. The velocity profiles are shown for different streamwise locations, X/R bq, starting from the exit plane of the burner quarl (X being the streamwise co-ordinate); the abscissa is the normalised radial co-ordinate Y/R bq, where R bq is the quarl radius, while the ordinate represents the axial velocity component in m/s. 1 X/Rbq= Y/Rbq X/Rbq= Y/Rbq X/Rbq= Y/Rbq X/Rbq= Y/Rbq X/Rbq=.1 1 r=.5 r=1.13 r= Y/Rbq X/Rbq=.1 r=.5 r=1.13 r= Y/Rbq Fig. 3: mean axial velocity profiles in the absence of air swirl motion, for different fuelair velocity ratio. Fig. : mean axial velocity profiles for air swirl number=1., for different fuel-air velocity ratio.

7 Examination of mean axial velocity profiles shown in Fig. 3 clearly puts into evidence the effect of fuel-air velocity ratio in the inner region, while the structure of the main co-flowing air stream is quite similar for the investigated operating conditions, as for velocity gradients, absolute values and radial extension. The presence of a velocity peak or lack in the central region indicates the residual structure of the fuel jet (with velocity values correspondent to that ones of nozzle injection), clearly noticeable until X/R bq = 1. Velocity lack for r=.5 is probably emphasized by wake effect downstream the injector, visible also for r=.9. At the quarl exit plane the spreading of the air stream is not enough to fill the quarl cross-section and a small recirculation appears close to the wall. The jet presents a quite slow radial expansion at progressive increasing distance from the efflux. Under swirling conditions (Fig. ), the radial profile of mean axial velocity reveals a well pronounced central recirculation zone (CRZ), characterised by negative values of the velocity, caused by the adverse pressure gradient induced by the intense swirl. Swirl presence ensures a rapid radial expansion of the flow with respect to no swirl condition, already at the first traverse downstream the quarl efflux. In this case too, the influence of the central fuel jet upon the recirculation region is clearly visible, especially at X/R bq =.1 position (for r=1.13, the CRZ is almost completely balanced by the central jet momentum, so that its intensity is highly decreased, remaining constant the radial extension). Velocity profile in the region external to fuel injection remains uninfluenced and is very similar for different values of r. Residual presence of the central jet disappears almost completely at X/R bq =.5, where the profiles are similar, independently from fuel-air velocity ratio (recirculation bubble develops also for r=1.13, although with a negative peak less pronounced). Integration of the axial velocity profiles made possible the evaluation of recirculation strength (defined as the ratio between the recirculated mass flow rate in the CRZ and the total injected flow rate through the burner quarl) for different operating conditions (Fig. 5). At X/R bq =.1 the recirculated air mass is highly influenced by the value of fuel-air velocity ratio (7% for r=.5; 1% for r=.9; 9% for r=1.13); the influence of this parameter progressively decreases far away from the efflux, so that at X/R bq =1, recirculation intensity assumes a quite constant value (about the 15% of the injected flow rate). 3 Mric/Mtot (%) 5 r =.5 r = 1.13 r = X/Rbq 1. Fig. 5: recirculation intensity for air swirl number=1., at progressive increasing distance from the quarl efflux. 7

8 The profiles of mean circumferential velocity measured by LDA for S = 1. (not reported here) show a concentrated vortex near the burner exit, with a core of solid-body rotation extending to about one fuel pipe radius from the axis. Downstream, the maximum tangential velocity occurs almost at the same radial distance from the axis as the peak of the axial velocity, showing the classical evolution from solid body rotation to Rankine type vortex, owing to turbulence dissipation effects, progressively going far away from the quarl region. Fig. reports a -D representation of the flow field, in a plane normal to the burner axis, at X/R bq =.5 and 1 measured by PIV (Particle Image Velocimetry) and averaged over 1 images. The -D analysis allows the estimation of radial velocity component, difficult to measure with LDV system owing to optical access. r =.5 X/Rbq =.5 Z mm Y mm r =.5 X/Rbq=1 Z mm Y mm Fig. : -D instantaneous flow field in a plane normal to burner axis for air swirl number=1. (reference vector length=7 m/s).

9 Turbulence intensity profiles (not shown here) indicate in general, for different operating conditions, low turbulence levels (about. V) inside the flow field with the highest values (.35 V) close to the positions of the highest velocity gradient in the inner shear layer of the annular air stream. In the case S = 1., it is observed that inside the CRZ region the rms levels are almost constant (.5 V) and do not decrease axially, in the investigated region. Fuel concentration distribution The fuel concentration distribution was measured by LSV, in a number of cross-sections normal to the burner axis, starting from the quarl exit, after collection of the images by a CCD camera located at 9 degree (Fig. ). The fuel line was seeded with oil droplets capable of tracing the gas diffusion into the main air stream. Owing to the quasi-monodisperse size distribution of the droplets, the scattered light intensity can be assumed proportional to their number density into the illuminated cross-section. Further, the particle concentration is assumed proportional to the fuel concentration, although the evaluation of the absolute value is a very difficult task [1]. The spatially averaged fuel concentration C at each cross-section, normalised by the initial mean concentration C at the fuel nozzle exit, was extracted from images as those shown in Fig. 7 (referred to S= and fuel-air velocity ratio=.5), taken by the planar visualisation system and representing the time averaged concentration field at progressive increasing distances form the efflux (traverse at X/R bq = -1 corresponds to fuel injection in the air stream). It was decided to present the results in terms of the quantity U, already defined, which is indicative of homogenisation of the fuel into the co-flowing air stream. The parameter U is independent on the initial absolute concentration of tracing particles, injected into the fuel, provided the initial average value C remains constant and the same integration area is used to evaluate the spatially averaged mean and rms concentrations. By comparing the values of U at different cross-sections, it is possible to evaluate the rate of homogenisation of the fuel in the co-flowing air stream [1]. The results reported in this paper are intended as a preliminary investigation on the reliability of this procedure to characterise the mixing efficiency and on the sensitivity to the effect of operating conditions. Fig. shows the evolution of unmixedness parameter for different operating conditions (S= - 1.; r= ). It is clearly noticeable that, for a fixed velocity ratio, the recirculating regime produced by the swirl presents the larger capability of an efficient and precocious mixing between the fuel jet and the main air stream. The reason is that, in this regime, the residence time of the outer stream incorporated in the recirculation zone is of the order of the mixing time, R bq /u, u being the rms velocity of the turbulent mixing layers. This residence time is approximately twice longer than the transit time, R bq /V, which characterises the flow without swirl (S = ). In fact, the absence of swirl leads to a less efficient homogenisation for the same distance. On the contrary, at a fixed swirl number, an increasing of velocity ratio r corresponds to a lower mixing efficiency, as already confirmed by LDV measurements, as for development of recirculation region and recirculated mass flow rate in the swirling case. The results indicate that the unmixedness index, U, may be a significant parameter to judge the mixing efficiency produced in a specific burner geometry by varying the swirl strength or the velocity ratio, at least under isothermal conditions. 9

10 X/R bq = -1 X/R bq =.1 X/R bq =.5 X/R bq = 1 Fig. 7: typical images of fuel distributions at increasing distances X/R bq from the quarl efflux for the following conditions: S=, r=.5..5 U. r =.5, S = r =.5, S = 1. r = 1.13, S = r = 1.13, S = X/Rbq Fig. : unmixedness parameter U as a function of different operating conditions. Conclusions The three dimensional gas velocity and fuel concentration field produced in the near-field turbulent mixing zone of a natural gas swirl burner have been investigated under isothermal conditions. The effect of swirl number and fuel-air velocity ratio has been quantified by 1

11 means of laser Doppler anemometry (LDA), particle image velocimetry (PIV) and laser sheet visualisation (LSV). The preliminary results indicate that the unmixedness index, U, may be a significant parameter to judge the mixing efficiency of a specific burner geometry, being sensitive to the swirl strength and the fuel-air velocity ratio, at least in the near field zone downstream the quarl exit. The method is based on the laser sheet visualisation technique and its reliability strongly depends on the quality of the planar images and the accuracy of normalisation of the measured average concentration by the mean concentration at the nozzle exit. Correlation with the turbulent field and shear stresses, measured by the LDA technique, is needed to complete the experimental investigation under isothermal conditions, before starting the combustion tests. The results obtained could be helpful for a more thorough comprehension of the reactants mixing process in model of industrial burners, providing useful informations for optimization of burner design and operating conditions, aiming to improve combustion efficiency and reduce environmental impact of fossil fuel combustion. Moreover, experimental data about the turbulent flow field can be used to validate results from numerical simulation through CFD codes. Aknowledgements The present experimental work has been performed at the CNR-TeMPE laboratories with partial support by MURST, under the national programme «Combustion». The authors would like to thank Mr. G. Brunello for the helpful support in the experiments and image analysis and Mr. C. Franzoni (Sensortech) for providing the PIV system. References 1. Chen, R.H. and Driscoll, J.F. (199) 3 rd Symposium (Intern.) on Combustion, The Combustion Institute, Pittsburgh, PA, pp.1-.. Cheng, T.S., Chao, Y.C., Wu, D.C., Yuan, T., Lu, C., Cheng, C. and Chang, J.M. (199) Effects of fuel-air mixing on flame structures and NOx emissions in swirling methane jet flames, 7 th Symposium (Intern.) on Combustion, Univ. of Colorado at Boulder, Co, August Lyons, V.J. (191) Fuel-air nonuniformity effect on nitric oxide emissions, AIAA J.,, No.5, p.. Fric, T.F. (1993) Effects of fuel-air unmixedness on NOx emissions, J. of Propulsion and Power, 9, No.5, p Tomeczek, J., Goral, J. and Gradon, B. (1995) Gasdynamic abatement of NOx emission from industrial natural gas jet diffusion flames, Comb. Sci. and Tech., 15, pp Hillemans, R., Lenze, B. and Leuckel, W. (19) Flame stabilisation and turbulent exchange in strongly swirling natural gas flames, 1 st Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp Chen, R.H. and Driscoll, J.F. (19), The role of the recirculation vortex in improving fuel-air mixing within swirling flames. nd Symposium (Intern.) on Combustion, The Combustion Institute, Pittsburgh, pp Charles, R.E., Emdee, J.L., Muzio, L.J. and Samuelsen, G.S. (19) The effect of inlet conditions on the performance and flowfield structure of a non-premixed swirl-stabilised distributed reaction, 1 st Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp Starner, S.H. and Bilger, R.W. (199) Further velocity measurements in a turbulent diffusion flame with moderate swirl, Comb. Sci. and Tech., 3, pp Shen, D., Most, J.M., Joulain, P. and Bachman, J.S. (199) The effect of initial conditions for 11

12 swirl turbulent diffusion flame with a straight-exit burner, Comb. Sci. and Tech., 1, pp Coghe. A. and Solero, G. (1997): Analisi del miscelamento turbolento aria/combustibile nel modello di un combustore a bassa produzione di NOx, Final Enea Report. 1. Solero, G., Brunello, G. and Coghe, A. (199) Effects of injection typology on turbulent mixing in the primary region of a homogeneous combustor, Combustion Meeting 9, Ravello, May, Gupta, A.K., Lilley, D.G. and Syred, N. (19) Swirl Flows, Abacus Press. 1. Birch, A.D., Brown, D.R., Dodson, M.G. and Thomas, J.R. (197) The turbulent concentration field of a methane jet, J. Fluid Mechanics,, Part 3, p Villermaux, E. (199) Mixing and Spray Formation in Coaxial Jets, J. of Propulsion and Power, 1, No.5, pp Chao, Y.C., Han, J.M. and Jeng, M.S. (199): A quantitative laser sheet image processing method for the study of the coherent structure of a circular jet flow, Experiments in Fluids, 9, pp