Centrifugal Mixing in Gas and Liquid Fuelled Lean Swirl Stabilized Primary Zones

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47 St., New York, N.Y The Society shall not be responsible for statements or opinions advanced in papers or in discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published in an ASME Journal. Papers are available from ASME for fifteen months after the meeting. Printed in USA. Copyright 1985 by ASME Centrifugal Mixing in Gas and Liquid Fuelled Lean Swirl Stabilized Primary Zones N. T. AHMAD, G. E. ANDREWS, M. KOWKABI and S. F. SHARIF Department of Fuel and Energy, University of Leeds, Leeds LS2.9JT, U.K. ABSTRACT A flat bladed axial swirler of 0.7 swirl number has been investigated in a 76 mm diameter combustor with all the combustion air passing through the swirler. Both liquid and gaseous fuels were used with eight injection points on the combustor wall just downstream of the swirler. This wall injection was aimed at the exploitation of centrifugal mixing forces acting on the burnt gas pockets at the wall to send them towards the centre and to displace higher density unburnt gas pockets to the wall and so promote mixing. For both kerosene and propane fuels there was a significant improvement in the combustion efficiency and NOX emissions compared with central fuel injection. For kerosene the NOX emissions were lower than for propane and very close to those for premixed fuel and air. However, for gas oil there was little improvement in performance with wall injection compared with central. This was attributed to the slower vaporisation rate with gas oil coupled with the centrifugal action on the liquid droplets with central injection. INTRODUCTION Many practical combustion devices employ swirling flow to accelerate the rate of flame spread and to enhance flame stability (1,2). Most practical swirl systems involve high swirl levels to generate a central recirculation zone and these are generally assumed to be necessary for good flame stabilisation. Most of the experimental bases for the advantageous properties of swirling flame come from studies of furnace burners or free flames (3,4). With these swirl burners central fuel injection is the only feasible fuel injection location. Also, the swirling jets spread rapidly in the radial direction with a rapid decrease in the swirl velocities. The application of swirling flow to gas turbines involves enclosure dimensions much closer to those of the swirler than encountered in furnace applications and hence swirl jet spread and velocity decay is limited. The resultant swirler aerodynamics may be considerably different from the furnace aerodynamics (5). For lean swirl based primary zones designs, of interest in the present work, the swirler has to be as large as possible relative to the combustor diameter if the pressure loss is not to be excessive. However, a major advantage of the enclosed gas turbine geometry is the possibility of injecting the fuel at the wall rather than at the centre. This may be with the swirl air inlet with radial swirlers (6) or as fixed injectors on the wall as in the present work. In swirl combustion centrifugal forces are important, although they have been ignored until recently (7-11). The local denisty differences, due to combustion of pockets of gas surrounded by unburnt gases, result in the lower density hot gases moving inwards relative to the high density unburnt gases. This relative movement promotes unburnt/burnt gas mixing and increases the turbulence levels and flame propagation rates. This effect will be maximised in enclosed swirl flows where the radial decay of the swirl velocity and hence centrifugal force is limited. For a furnace application the effect would only be significant very close to the burner. For propane the authors (12) have shown that for an enclosed swirl flow there was a major difference in the combustion and emissions performance between central and wall fuel injection. Wall injection had a much better combustion efficiency and lower NOX emissions than for central injection. This was attributed to the influence of centrifugal mixing forces with wall injection. Internal gas analyses traverses were carried out (12) which showed that with central injection the fuel remained in the central core region and the flame had great difficulty in propagating outside this region. With wall injection the fuel was very quickly uniformly distributed and the radial profiles downstream were close to those for premixed combustion. In the present work the extension of this previous work to liquid fuels is investigated as well as reporting further details of the propane studies. Ahmad and Andrews (13) have investigated the performance of the present swirler using central injection of liquid fuels and compared the results with central propane injection. They showed that there was Presented at the 1985 Beijing International Gas Turbine Symposium and Exposition Beijing, People's Republic of China September 1-7, 1985

2 TABLE 1: SW3 Swirler Design Details Blade Type :Flat Hub Diameter :14.8mm Swirler Outer Diameter, d :47.8mm Combustor diameter, D :76mm D/d :1.59 Swirl number (15) :0.71 Pressure Loss at Combustor Mach Number of 0.047:4.7.; Blockage :771 flow aerodynamics. For liquid fuels the hole size was such as to achieve good air assisted atomisation (16). FIG. 1: The Swirl Combustor Geomegry 8 WALL FUEL HOLES a major difference in the combustion performance with liquid fuels. The weak extinction occurred at a much richer mixture than for propane and this prevented a high combustion efficiency and low NOX emissions from being achieved. It was argued that a possible explanation for this poor combustion performance could be the action of centrifugal forces on the liquid droplets sending them out into the wall region so that the fuel placement was completely different to that of propane. In the present work direct injection of liquid fuel at the wall enables this hypothesis to be tested. EXPERIMENTAL EQUIPMENT Axial Swirler Design The aim of improving the fuel and air mixing with fuel injection at the wall was to achieve a situation closer to that of premixed situations so that low NOX emissions as well as a high combustion efficiency could be achieved. Ahmad and Andrews (14) investigated the performance of a range of swirlers of different swirl number and pressure loss with premixed combustion. They showed that for weak mixtures a high swirl number was detrimental to the achievement of a high combustion efficiency. Also an increase in the swirler pressure loss at approximately constant swirl number resulted in an improvement in combustion efficiency with no significant increase in NOX. From this work the SW3 swirler was chosen for the present investigations and those for central fuel injection (13) as it had a realistic pressure loss of 4.7% at typical gas turbine operating conditions. The full swirler design details are given in Table 1 and the combustor configuration is detailed in Fig. 1. The swirl number was sufficient to create a strong central recirculation zone. The size of this was determined from axial wall static pressure profiles and was approximately 150 mm downstream of the swirler. Fuel Injector Design For propane injection eight equispaced fuel holes were used of diameter 1.3 mm. For liquid fuels an equivalent geometry was used with a fuel hole size of 0.3 mm. These fuel orifice sizes gave fuel injection velocities of m/s for propane and 5-8 m/s for liquid fuels, depending on the equivalence ratio used. The propane velocities were similar to the air velocity in the swirler vane passages which was approximately 100 m/s and hence would tend to follow the air With central fuel injection a 12.7 mm diameter fuel pipe was used with the fuel hole approximately 10 mm downstream of the swirler. Two fuel injectors were used, one for propane and one for liquid fuels. For wall injection a 76 mm diameter combustor insert 25 mm long was used which carried both the eight 1.3 mm diameter propane fuel injection points and the eight liquid fuel injection holes. Two fuel manifolds were welded to the two sets of eight equispaced fuel injectors. The liquid fuel injection points were offset half a pitch from the propane injection points but in the same axial plane approximately 10 mm downstream of the swirler. To reduce the possibility of the fuel being swept downstream the fuel injection holes were inclined 30 degrees upstream. A further advantage of the wall injection was the better initial spatial distribution of the fuel which would result in more rapid fuel and air mixing. Combustor Test Facility The 76 mm diameter atmospheric pressure combustion test facility has been described previously (12-14). It consists of an air supply from a fan, venturi flow metering, electrical air preheaters, 1.5m long 76 mm diameter approach pipe, swirler and fuel injector, 330 mm long 76 mm diameter uncooled combustor and an exhaust system with a flame observation window on the combustor centreline. Ignition was achieved with a 12 Joule surface discharge ignitor mounted 25 mm downstream of the wall fuel injector section. Combustor Test Conditions The air was supplied at a Mach number based on the 76 mm combustor diameter of which is typical of current gas turbine combustors. Electrical preheat was used to raise the air temperature to 600K, which is typical of industrial gas turbine high power conditions and the approach condition in aero gas turbines. At this inlet temperature the mean air velocity was 24 m/s. For a gaseous fuel industrial propane was used and its composition was determined by gas chromatography and incorporated in the gas analyses computer programme. For liquid fuels the hydrogen content is the most important parameter from a combustion viewpoint. This was determined using an NMR technique and was 13.76% for the kerosene fuel and 13.04% for the gas oil fuel. A detailed analysis of the two fuels for PAH compounds which can influence smoke and NOX emission is presented in Ref. 17. Gas Sampling System Mean gas samples were obtained at the combustor exit plane using a stainless steel water cooled 'X' probe with twenty holes on centres of equal area. The sample gas outlet temperature from the probe was 2.

3 X w 5 r/rx 1Y A r/r X 1 CT- 1 A V + x FIG. 2: Radial Adiabatic Flame Temperature Profiles monitored and kept above 150 C by regulating the water flow. The sample was transported along an electrically heated (150 C) teflon line to a heated sample pump and filter housed in an oven. The sample line was purged with heated clean air during the ignition process. Automatic high temperature solenoid valves were incorporated into the oven to switch from sampling to purge to calibration gases. This heated sampling system ensured that no unburnt liquid fuel could condense out prior to arrival of the sample at the heated FID hydrocarbon analyser. Some radial traverses were also obtained using the single point water cooled probe described in Ref. 12 and the same heated sample line system. WEAK EXTINCTION The weak extinction equivalence ratio was measured at a constant air mass flow rate by gradually reducing the fuel flow until the flame was extinguished. The process was observed directly from the control room through the 100 mm diameter observation window on the combustor centre line. Also weak extinction was monitored from the gas analysis console by a sudden increase in the hydrocarbon emissions. Weak extinction was repeatable to within 0.02 of an equivalence ratio. The measured weak extinction data are summarised in Table 2 which compares the present results for wall injection with those in Ref. 13 for central injection. TABLE 2: Weak Extinction (600K and M=0.047 Fuel Propane Kerosene Gas Oil Premixed Propane Central Injection Wall Injection (Ref. 13) With wall injection of fuel it was anticipated that the enhanced mixing between pockets of burnt and unburnt gases due to the action of centrifugal forces would lead to a reduction in the flame stability compared with central injection of the fuel and a weak extinction value closer to that for premixed fuel and air. Table 2 shows that this only occurred for propane. For kerosene there was little change in the weak extinction with wall injection, whereas for gas oil there was a major extension of the flame stability and a FIG. 3: Radial Variation of Equivalence Ratio superior stability to that of propane with central or wall injection. The greater stability range with gas oil relative to kerosene was expected, due to the slower vaporisation rate and hence creation of locally richer zones. However, by the same reasoning a wider stability range with kerosene than for propane would be expected and this was not observed, either for central or wall injection. Clearly the method of flame stabilisation for all three fuels must be different if these weak extinction observations are to be explained. Internal gas composition traverses, similar to those reported in Ref. 12 for propane, have been carried out with the present liquid fuel injection at the wall. These will be reported in detail elsewhere, but they do show differences between the three fuels which may explain the stability differences. For propane the flame was stabilised for both central and wall injection by combustion in the shear layer at the boundary of the inner core recirculation zone and the outer swirling flow. The internal traverses 30 mm from the swirler showed a local high temperature zone in this region. With kerosene this was absent and the temperature profile at the equivalent position was very close to that for premixed combustion (12), as shown in Fig. 2. It may therefore be possible that for kerosene the flame was stabilised by vaporised fuel recirculated into the core region, by which stage it will be well mixed with the air and hence burn as a premixed mixture. The flame stability mechanism will then be the same as for the premixed flame and will be governed by the proportion of the kerosene which is recirculated into the core region. The internal composition profiles in Fig. 3 show that the local equivalence ratio of the inner core region with wall injection of kerosene is close to that of the premixed weak extinction. This indicates that with kerosene the flame is extinguished when the inner core region equivalence ratio falls below that for premixed weak extinction (0.41). The fact that this equivalence ratio was weaker than that for the overall kerosene weak extinction merely indicates that not all of the kerosene was vaporised by the end of the recirculation zone. Slow vaporisation of the kerosene in the outer swirling air, which remains at the inlet temperature during the recirculation zone length, is thus a key to the poor flame stability. The action of centrifugal forces in this situation is twofold: firstly, in the initial outer swirl flow any vaporised fuel with a density greater than that of air will be forced to remain in the wall region, secondly, when 3.

4 X102 X102 a I w 2 3 AXIAL DISTANCE X 1.02 A V X EQR FIG. 4: SW3W Kerosene Wall Temperature Profiles, 600K the combustion has been initiated in the outer swirl flow, downstream of the recirculation zone, centrifugal mixing of burnt and unburnt gases enhances the mixing and rates of flame spread. The mean exhaust plane gas composition results will be shown to support this hypothesis of the combustion mechanism for wall injection of kerosene. With gas oil as the fuel a similar mechanism is likely to occur. However, due to the slower vaporisation rates there will be less fuel in the vapour phase at the end of the recirculation zone and hence weaker mixtures recirculated into the core region. The internal gas composition profiles in the recirculation zone shown in Fig. 3 for an overall equivalence ratio of 0.6 show equivalence ratios weaker than for kerosene and outside the weak extinction for premixed propane/ air. Thus the flame cannot be stabilised by the recirculation zone shear layer. A possible stabilisation mechanism may be the flow expansion downstream of the recirculation zone where the outer swirl flow decelerates and turbulence is generated. Wall static pressure profiles show that this occurs approximately 150 mm downstream of the swirler (5). The internal gas composition profiles at this point showed that much of the fuel vaporisation occurs in this region and the rich zones that are generated enable very weak overall mixtures to be stabilised. Again the mean exhaust plane gas composition measurements reported below support this hypothesis for the combustion mechanism with wall injection of gas oil. In this situation, centrifugal forces will have a similar effect on mixing as discussed above for kerosene. The importance of fuel vaporisation in the outer swirling air flow at the inlet temperature is supported by the inability to stabilise any flame at the 400K inlet temperature for both kerosene and gas oil, even if a premixed flame was first established. In Ref. 13 it was postulated that for central injection of liquid fuels the centrifugal force could act on the droplets to send them into the wall region and hence the combustion performance could be similar to that with wall injection. Table 2 shows that in terms of weak extinction this is a valid hypothesis for kerosene. However, for gas oil the vaporisation may be sufficiently greater than that for wall injection, due perhaps to better atomisation with central injection, for the central recirculation zone to be richer than for wall injection and to thus stabilise the flame through the premixed core mechanism discussed for 2 3 AXIAL DISTANCE X 102 A V + X 0 FOR FIG. 5: SW3W Gas Oil Wall Temperature Profiles, 600K kerosene. COMBUSTOR WALL TEMPERATURE AXIAL DEVELOPMENT The combustor wall was uncooled and its temperature was monitored by an axial row of grounded junction mineral insulated Type K thermocouples. The location of the heat release zone was easily detected by an increase in the wall temperature. Fig. 2 shows that at the 50 mm axial position the flame had not propagated to the wall. As the bulk of the air mass flow is in this outer unburnt region, the overall heat release does not occur until the fuel in this region burns. This results in the increase in the wall temperature corresponding with the main heat release zone. The temperature profiles for a range of equivalence ratios are shown in Fig. 4 for kerosene and Fig. 5 for gas oil, as a function of the axial distance downstream of the swirler exit plane. The equivalent profiles for central injection were given in Ref. 13. Fig. 6 shows for an equivalence ratio of 0.6, the comparison of the wall temperature profiles for premixed combustion and for wall injection with all three fuels. Figs. 4 and 5 show similarities in the temperature profiles for the two liquid fuels. However, for the same equivalence ratio the kerosene wall temperatures are higher in the initial region but lower in the region downstream of the recirculation zone. For gas oil in the initial region the wall temperatures are significantly lower than the air inlet temperature, indicating possible liquid fuel cooling of the wall in this area. Both fuels show little influence of the equivalence ratio on the wall temperature in the initial region. This supports the flame stabilisation mechanism discussed above which involves fuel vaporisation taking place in the wall region throughout the recirculation zone length. The higher wall temperatures in the downstream region of the combustor supports the above mechanism of the gas oil flame stabilised by local rich zones in the wall region just downstream of the recirculation zone. Fig. 6 shows that for premixed and propane wall injection the flame is stabilised much closer to the swirler than for wall injection of kerosene and gas oil. The lower temperatures for kerosene are due to the closeness of the equivalence ratio used to the weak extinction. However, Fig. 4 shows that at richer mixtures kerosene would achieve a high temperature in the downstream portion of the flame tube. Fig. 6 4.

5 X102 U 2 3 AXIAL DISTANCE X 102 A V FIG. 6: SW3W Wall Temperature Profiles, EQR=0.6 demonstrates that the vaporisation period for liquid fuel delays the flame development as discussed above, in relation to the weak extinction. The zone of rapid temperature rise and hence of heat release in the wall region is approximately 50 mm downstream of that for propane. CARBON MONOXIDE EMISSIONS The mean exhaust carbon monoxide emissions for wall injection of propane, kerosene and gas oil are shown as a function of equivalence ratio in Fig. 7. The equilibrium CO levels are also shown to indicate the proportion of the measured CO that could be attributed to equilibrium CO. An equivalent combustion 'inefficiency' will be referred to later in the combustion inefficiency results. For propane the results are close to those for premixed combustion (14) indicating that the wall injected fuel has achieved good mixing with the swirling air with no rich zones to generate high levels of CO. This was supported by the internal gas composition traverses (12) which showed radial CO profiles only slightly above those for premixed combustion, whereas for central injection there was a rich central core region with very high CO levels. It is considered that the action of centrifugal mixing forces, in retarding mixing with central propane injection and accelerating it with wall injection, was responsible for both the local internal and the mean exhaust plane CO levels for wall injection being close to those for premixed combustion. Fig. 7 shows that for wall injection of kerosene the CO emissions are close to those for propane injection and lower than for propane for equivalence ratios richer than In this region the kerosene CO emissions are very close to the equilibrium levels. Only the poor flame stability prevented the achievement of very low CO emissions with kerosene. Comparison of the present results with those for central injection of kerosene in Ref. 13, shows that the CO levels were much lower with wall injection by a factor which increases from approximately 1.5 at equivalence ratios near weak extinction to over 4 at richer mixtures. The observation of low CO levels with wall injection of kerosene supports the combustion mechanism postulated in the discussion of the weak extinction results. Recirculation of vaporised kerosene to the FIG. 7: Variation of CO Concentration with EQR, Ref.M, 600K central core region produces a well mixed mixture there which both stabilises the flame and ensures that there are no locally rich regions which generate high CO emissions. This model requires that the kerosene results with wall injection are similar to those for premixed combustion and this is demonstrated in the present CO emissions and also in the internal radial gas composition profiles. The slow vaporisation of the kerosene allows time for mixing as the vaporisation proceeds. With propane injection, combustion can occur earlier than for kerosene and the mixing will be less complete when combustion occurs, and hence CO emissions higher at the same equivalence ratio. The CO emissions for wall injection of gas oil are shown in Fig. 7 to be an order of magnitude higher than for kerosene. Comparison with the results for central injection of gas oil in Ref. 13 shows that the present CO emissions are similar but with slightly higher levels at all equivalence ratios. This indicates that the combustion mechanism is likely to be similar in both cases, so that with central injection the centrifugal forces send the droplets into the wall region. As discussed in relation to the weak extinction results the slower vaporisation rate will prevent the formation of fuel vapour-air mixtures that are flammable by the end of the recirculation zone and the flame stabilises in the rich wall region as the fuel vaporises in the flow expansion downstream of the recirculation zone. With wall injection it is probable that the swirling air velocities will be lower than at the swirler vane passage exit and hence the drop size will be greater (16). This will lead to a slower vaporisation rate with wall injection and hence to the observed locally richer wall mixtures and CO emissions. It is possible that the present atmospheric pressure results could be considerably modified for gas oil at engine pressure due to the decrease in drop size and increase in vaporisation rates with pressure. The situation observed in the present work for kerosene may then apply and the very poor combustion performance for gas oil could then be significantly improved. UNBURNT HYDROCARBON EMISSIONS The mean exhaust plane unburnt hydrocarbon (UHC) emissions are shown as a function of equivalence ratio in Fig. 8. With wall injection of propane the UHC emissions are very close to those for premixed combustion. Again this is a reflection of the enhanced 5.

6 LOG 10 1 LOG 10 U I 4 6 EaRX10 A V x FIG. 8: Variation of UHC with EQR at 600K centrifugal mixing with wall injection. With central injection of propane the UHC emissions were an order of magnitude higher due to the excessively rich central core zone (12,13). Internal gas composition traverses also showed a very close similarity with those for premixed combustion with very low UHC levels in the central core region and higher levels in the wall region (12). With a heavier than air gaseous fuel injected at the wall fuel cannot reach the centre unless it has mixed with air and been burnt. With a liquid fuel injected at the wall the vaporisation period alters this situation. It has been argued for kerosene that this period results in most of the fuel being vaporised by the end of the recirculation zone. Recirculation back into the central core recirculation zone of part of the vaporised fuel with the air that it has mixed with during the vaporisation period, results in premixed combustion in the core. The UHC emissions for kerosene shown in Fig. 6 support this argument with levels close to those for premixed combustion at equivalence ratios richer than 0.7. The increase in UHC emissions at weaker mixtures is due to the narrower flame stability limits discussed above. For wall injection of gas oil, Fig. 8 shows that the UHC emission levels are considerably above those for kerosene and propane. However, the levels are not excessively high in the equivalence ratio range 0.4 to 0.6 and are similar to those for central injection of propane. The longer vaporisation period with gas oil would be expected to lead to a higher level of UHC emissions than for kerosene, for the same overall residence time. It has been argued above that the slower vaporisation with gas oil results in insufficient fuel recirculated into the central core region for combustion to occur there. The resultant rich burning zones in the outer swirl region downstream of the recirculation zone have a sufficient temperature to convert most of the vaporised fuel to the CO stage of the combustion process but insufficient time to burn out the high CO levels generated by these rich zones. It is possible that with additional combustor length the high CO levels discussed above could be considerably reduced. The present combustor length is a minimum size based on aero applications and for industrial applications could be much longer and the performance with gas oil could be considerably improved. 6 EGA X10` 1 A V x FIG. 9: Variation of Combustion Inefficiency with EQR at 600K COMBUSTION INEFFICIENCY The combined influence of CO and UHC emissions are presented in terms of the combustion inefficiency as a function of equivalence ratio in Fig. 9. The results reflect the trends discussed above for CO and UHC emissions. Premixed and wall injection of propane and kerosene all gave a low combustion inefficiency, whereas for gas oil the much higher inefficiency represents unacceptable levels. This high inefficiency for gas oil is mainly due to the high CO emissions. NOX EMISSIONS The NOX emissions as a function of the mean flame temperature, computed from the mean gas composition, are shown in Fig. 10. The NOX emissions corrected to 15% oxygen are shown as a function of equivalence ratio in Fig. 11. With propane injected at the wall the NOX levels are a little higher than for premixed combustion. However, when compared with the NOX emissions for central propane injection (12,13) there is a considerable reduction. This is due to the enhanced mixing due to the centrifugal influence. With kerosene injected at the wall the NOX emissions were very close to those for premixed combustion as shown in Figs. 10 and 11. These low NOX levels, well below those for propane, support the proposed combustion mechanism of recirculation of vaporised kerosene mixed with air into the central core region where it burns as a premixed mixture with the observed low NOX emissions as a consequence. The internal radial gas composition profiles also show lower NOX emissions than for propane and gas oil at all radial positions. This result represents the lowest NOX emissions for any gas turbine combustion system burning kerosene without the necessity to prevaporise and premix upstream of the combustor. However, as these processes occur internally in the swirl combustor the problem of a poor flame stability remains, as for all premixed systems. The NOX emissions for wall injection of gas oil are dramatically different to those for kerosene and are at a very high level at all equivalence ratios except those just prior to weak extinction. Again the causes are directly related to those which result in the much wider stability limits than for kerosene. The vaporisation rate is too slow to achieve sufficient fuel recirculation into the core region to achieve a 6.

7 LOG10-1 LOG10 O it X FLAME TEMP (K) X 102 o V PROPANE KEROSENE GASOIL FIG. 10: NOX as a Function of Flame Temperature flammable mixture. The resultant flame downstream of the recirculation zone has rich regions in the wall region which generate the high NOX emissions. Internal gas composition measurements have shown that the high levels of NOX are mainly generated in the last third of the combustor length where the wall temperature profiles in Fig. 3 shows the main heat release to occur. At other operating conditions such as higher inlet temperatures or higher pressures that will increase the vaporisation rate, it is possible that gas oil will change its combustion mode to that of kerosene with a consequent improvement in the NOX emissions. COMBUSTION INEFFICIENCY AND NOX Low NOX emissions are only useful if they are achieved in combination with a high combustion efficiency. An optimum operating equivalence ratio can be determined by plotting the combustion inefficiency against the NOX emissions corrected to 15% oxygen. The present results are plotted on this basis in Fig. 12. This shows that with kerosene there was a minimum combustion inefficiency with NOX emissions close to those for propane. The gas oil results show that a satisfactory combustion inefficiency was never achieved and the NOX emissions were always unacceptably high. An optimum primary zone operating condition may be established by taking the NOX emissions at the point just before the combustion inefficiency increases, on the low NOX side of Fig. 12, and obtaining the equivalence ratio at which this NOX emission was achieved from Fig. 11. These conditions for the present results are summarised in Table 3 which also includes the results for premixed (14) and central fuel injection (13) Table 3 shows that for propane the optimum equivalence ratio and hence flame temperature is approximately the same whether the fuel is injected centrally or at the wall or is premixed. Only the NOX emissions and the combustion inefficiency show a sensitivity to the method of fuel injection. For kerosene the optimum equivalence ratio is richer with wall injection than for central and yet the NOX emissions are considerably reduced. This is a clear indication of a drastically different combustion mechanism as has been discussed. However, for gas oil the optimum equivalence ratio and NOX emissions are very similar for the two modes of fuel injection indicating that the combustion mechanism may be similar as postulated in Ref. 13. The only previous investigation of swirling flames 6 EGRX10-1 A V FIG. 11: Variation TABLE 3: Optimum of Corrected NOX with EQR at 600K Primary Zone Conditions at 600K and NOX ppm Equivalence Temperature 15% Oxygen Ratio Premixed Wall Propane Wall Kerosene Wall Gas Oil Central Propane Central Kerosene Central Gas Oil with fuel injection at the wall is that of Mestre (7, 18). This was aimed at the exploitation of the centrifugal mixing forces for a kerosene fuelled system. However, it was applied to an annular swirl flow with the inside cylinder water cooled and was a rather unrealistic situation for gas turbines. However, the results (18) included internal gas composition measurements for both swirling and non-swirling flows. These showed a significant improvement in the combustion efficiency and a reduction in the NOX levels with wall injected fuel and swirling flow compared with non-swirling flow. This was attributed to the beneficial influence of centrifugal mixing forces and better atomisation with the swirling flow. Work by one of the present authors (14,16,19) has shown that non-swirling flow grid plate type combustion systems can achieve emissions levels that are better than the present swirling flow results except for kerosene injection at the wall. Thus we do not agree with the conclusions of Mestre regarding non-swirling systems. The use of fuel injection at the wall has not been investigated previously for practical enclosed axial vane swirler stabilised flames. The technique is particularly suited to gas turbine combustor applications as the swirler expansion (d/d) is limited to less than two and hence the swirl velocities in the wall region remain sufficiently high to atomise the liquid fuel. The CIVIC combustor design of Shekleton (8,9) was aimed at the exploitation of the centrifugal mixing effect. However, the fuel was injected closer to the centre than the wall and the results correspond closely to those for central injection in the present work (12), particularly in terms of NOX emissions. Vranos et al (10) have also recognised the possibilities of enhancing flame propagation rates by initiating combustion at the periphery of a swirl flow, but only a low swirl was investigated and no detailed gas composition measurements were taken. 7.

8 LOG10 2. Syred, N. and Beer, J.M., 'Combustion in swirling flows - a review', Combustion and Flame, Vol. 22, 1974, pp.143. Ui z 3. Claypole, T.C. and Syred, N., 'The effect of swirl burner aerodynamics on NOX formation', Eighteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1981, pp Beltagui, S.A. and Maccallum, N.R.L., 'Aerodynamics of vane-swirled flames in furnaces', J.Inst. Fuel, Vol. 49, 1976, p.183. A PROPANE KEROSENE GASOIL 4 6 NOXC% X10 3 FIG. 12: Variation of Combustion Inefficiency with Corrected NOX Emissions at 600K CONCLUSIONS 1. In enclosed swirling flow combustion centrifugal forces have an important influence on flame propagation and fuel/air mixing rates. These effects are only present once combustion has been initiated. For propane they are present in the initial region downstream of the swirler. However, for liquid fuels the vaporisation period dealays the onset of combustion and the centrifugal mixing effects are confined to the latter half of the combustor. 2. For propane wall injection compared with central injection results in a decrease in the stability limits an improvement in the combustion inefficiency and a reduction in the NOX emissions. All these factors may be attributed to the improved mixing with wall injection. 3. With kerosene wall injection the weak extinction and emissions results all support a combustion mechanism based on recirculation of vaporised kerosene into the core region by which time it will be well mixed with the recirculated air. NOX emissions were close to the premixed situation and represent the lowest reported NOX emissions for a directly fuelled combustor design. 4. With gas oil wall injection the performance was very similar to that with central injection and none of the favourable emissions characteristics for kerosene wall injection were found. It was concluded that the vaporisation rate was too slow to establish a flammable mixture in the core region and the flame was established in rich wall regions downstream of the recirculation zone. ACKNOWLEDGEMENTS We would like to thank the U.K. Science and Engineering Research Council for a research grant (GR/B/ 44668) in support of the central injection part of this work. The test facility was operated by R.A. Boreham. REFERENCES 1. Gupta, A.K., Lilley, D.G. and Syred, N., 'Swirl Flows', Abacus Press, Ahmad, N.T. and Andrews, G.E., 'Enclosed axial vane swirler stabilised premixed flames'. Submitted for publication. 6. Roberts, P.B., Kubasco, A.J. and Sekas, N.J., 'Development of a low NOX lean premixed annular vortex combustor for an industrial gas turbine', ASME Paper 80-GT-58, Mestre, A. and Benoit, A., 'Combustion in swirling flow', Fourteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1972, pp Shekleton, J.R., 'The CIVIC - a concept in vortex induced combustion,' Trans. ASME, J. Eng. Power, Vol. 103, 1981, p Shekleton, J.R., 'The CIVIC - a concept in vortex induced combustion - Part 2', Trans. ASME, J.Eng. Power, Vol. 103, p Vranos, A., Knight, B.A. and Zabielski, M.F., Centrifugal Mixing: A comparison of temperature profiles in non-recirculating swirling and nonswirling flames', Combustion and Flame, Vol. 48, 1982, p Markowski, S.J., Lohmann, R.P. and Reilly, R.G., 'The vortex burner - A new approach to gas turbine combustors', Trans. ASME, J. Eng. Power, Vol. 98, 1976, p Ahmad, N.T., Andrews, G.E., Kowkabi, M. and Sharif, S.F., 'Centrifugal mixing forces in enclosed swirl flames', Twentieth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, Ahmad, N.T. and Andrews, G.E., 'Gas and liquid fuel injection into an enclosed swirling flow', ASME Paper 84-GT-98, Ahmad, N.T. and Andrews, G.E., 'Emissions from enclosed swirl stabilised premixed flames', ASME Paper 83-GT-192, Kerr, N.M. and Fraser, D., 'Swirl, Part 1: Effect on axisymmetrical turbulent jets', J.Inst.Fuel, Vol. 38, 1965, pp Abdul-Aziz, M.M. and Andrews, G.E., 'The influence of flame stabiliser pressure loss on fuel atomisation, mixing and combustion performance', Paper to be presented at the Third International Conference on Liquid Atomisation and Spray Systems, London, July,

9 17. Williams, P.T., Bartle, K.D., Andrews, G.E. and Mills, D.G., 'Capillary column G.C. analysis of Diesel fuels using simultaneous parallel triple detection', Sixth International Symposium on Capillary Column Chromotography, Italy, May, Mestre, A., 'Efficiency and air pollution formation studies in a swirling flow combustor', Proc. ASME Fluid Mech. of Combustion Conf., Montreal, 1974, p Andrews, G.E., Abdul-Aziz, M.M. and Al-Dabbagh, N.A., 'Mixing and fuel atomisation effects on premixed combustion performance', ASME Paper, 83- GT