The effect of hydrogen containing fuel blends upon ﬂashback in swirl burners

Lean premixed swirl combustion is widely used in gas turbines and many other combustion Processes due to the beneﬁts of good ﬂame stability and blow off limits coupled with low NO x emissions. Although ﬂashback is not generally a problem with natural gas combustion, there are some reports of ﬂashback damage with existing gas turbines, whilst hydrogen enriched fuel blends, especially those derived from gasiﬁcation of coal and/or biomass/industrial processes such as steel making, cause concerns in this area. Thus, this paper describes a practical experimental approach to study and reduce the effect of ﬂashback in a compact design of generic swirl burner representative of many systems. A range of different fuel blends are investigated for ﬂashback and blow off limits; these fuel mixes include methane, methane/hydrogen blends, pure hydrogen and coke oven gas. Swirl number effects are investigated by varying the number of inlets or the conﬁguration of the inlets. The well known Lewis and von Elbe critical boundary velocity gradient expression is used to characterise ﬂashback and enable comparison to be made with other available data. Two ﬂashback phenomena are encountered here. The ﬁrst one at lower swirl numbers involves ﬂash- back through the outer wall boundary layer where the crucial parameter is the critical boundary velocity gradient, G f . Values of G f are of similar magnitude to those reported by Lewis and von Elbe for laminar ﬂow conditions, and it is recognised that under the turbulent ﬂow conditions pertaining here actual gradients in the thin swirl ﬂow boundary layer are much higher than occur under laminar ﬂow conditions. At higher swirl numbers the central recirculation zone (CRZ) becomes enlarged and extends backwards over the fuel injector to the burner baseplate and causes ﬂashback to occur earlier at higher velocities. This extension of the CRZ is complex, being governed by swirl number, equivalence ratio and Reynolds Number. Under these conditions ﬂashback occurs when the cylindrical ﬂame front surrounding the CRZ rap- idly accelerates outwards to the tangential inlets and beyond, especially with hydrogen containing fuel mixes. Conversely at lower swirl numbers with a modiﬁed exhaust geometry, hence restricted CRZ, ﬂash- back occurs through the outer thin boundary layer at much lower ﬂow rates when the hydrogen content of the fuel mix does not exceed 30%. The work demonstrates that it is possible to run premixed swirl burners with a wide range of hydrogen fuel blends so as to substantially minimise ﬂashback behaviour, thus permitting wider used of the technology to reduce NOx emissions. (cid:2) 2011 Elsevier Ltd. All rights reserved.


a b s t r a c t
Lean premixed swirl combustion is widely used in gas turbines and many other combustion Processes due to the benefits of good flame stability and blow off limits coupled with low NO x emissions. Although flashback is not generally a problem with natural gas combustion, there are some reports of flashback damage with existing gas turbines, whilst hydrogen enriched fuel blends, especially those derived from gasification of coal and/or biomass/industrial processes such as steel making, cause concerns in this area. Thus, this paper describes a practical experimental approach to study and reduce the effect of flashback in a compact design of generic swirl burner representative of many systems. A range of different fuel blends are investigated for flashback and blow off limits; these fuel mixes include methane, methane/hydrogen blends, pure hydrogen and coke oven gas. Swirl number effects are investigated by varying the number of inlets or the configuration of the inlets. The well known Lewis and von Elbe critical boundary velocity gradient expression is used to characterise flashback and enable comparison to be made with other available data.
Two flashback phenomena are encountered here. The first one at lower swirl numbers involves flashback through the outer wall boundary layer where the crucial parameter is the critical boundary velocity gradient, G f . Values of G f are of similar magnitude to those reported by Lewis and von Elbe for laminar flow conditions, and it is recognised that under the turbulent flow conditions pertaining here actual gradients in the thin swirl flow boundary layer are much higher than occur under laminar flow conditions. At higher swirl numbers the central recirculation zone (CRZ) becomes enlarged and extends backwards over the fuel injector to the burner baseplate and causes flashback to occur earlier at higher velocities. This extension of the CRZ is complex, being governed by swirl number, equivalence ratio and Reynolds Number. Under these conditions flashback occurs when the cylindrical flame front surrounding the CRZ rapidly accelerates outwards to the tangential inlets and beyond, especially with hydrogen containing fuel mixes. Conversely at lower swirl numbers with a modified exhaust geometry, hence restricted CRZ, flashback occurs through the outer thin boundary layer at much lower flow rates when the hydrogen content of the fuel mix does not exceed 30%. The work demonstrates that it is possible to run premixed swirl burners with a wide range of hydrogen fuel blends so as to substantially minimise flashback behaviour, thus permitting wider used of the technology to reduce NOx emissions.

Introduction
Lean premixed (LP) combustion is a widely used strategy to decrease undesirable emissions in gas turbines. In LP systems, fuel and air are mixed prior to the combustion chamber to promote mixing, combustion efficiency, uniform temperatures and low NO x . Swirl combustors are almost universally used in some form or other in gas turbine [1][2][3] and numerous other systems. Especially when operated in a LP mode many problems can be encountered including blow off and flashback [2][3][4].
Using alternative fuels has become another option to reduce emissions of CO 2 . Hydrogen, hydrogen and other fuel blends can cause major issues with many swirl combustors, because of the considerable variation in flame speed with such fuel blends compared to natural gas. Similar comments apply to process gases such as coke oven gas (COG) widely produced in the steel industry. Biomass and coal gasification prototype power plants have performed well, but have not proved to be competitive against conventional boiler technology for power production [5][6][7], primarily because gas turbine manufacturers have had full order books for conventional units. Demand for systems capable of economically and efficiently producing power and CO 2 for sequestration may well change this. There are many other problems associated with the use of alternative fuels as discussed in [8].
Basically, swirling flows are defined as a flow undergoing simultaneous axial-tangential vortex motion. This flow motion can be generated using swirl vanes or many other methods [9,10]. The 0306-2619/$ -see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.apenergy.2011.01.057 main desirable characteristic of swirl combustors is the formation of unattached reverse flow zones (RFZ) and central recirculation zones (CRZ) capable of recycling hot chemically active reactants to substantially enhance flame stability [4]. The swirl number (S) is one of the main parameters used to characterise swirling flow. It is defined as the ratio of axial flux of swirl momentum divided by axial flux of axial momentum, divided by the equivalent nozzle radius [3]. Commonly owing to flow complexities a geometric swirl number (S g ) is used which depends entirely on the geometry of the burner.
Flashback is a problem which has arisen when using LP combustors especially with hydrogen based fuel mixtures. Flashback occurs when the gas velocity becomes lower than the burning velocity due to flame propagation within boundary layer, core flow or because of combustion instabilities [2,[11][12][13]. One important manifestation of the flashback phenomenon is that due to flame propagation in the low velocity region of the wall boundary layer. Flame propagation is thus limited by quenching in the very near wall region [13]; for turbulent flow this will be the laminar sub layer. Lewis and von Elbe [14] have suggested use of the critical boundary velocity gradient, based on considerations of the velocity gradient G f at the wall, the laminar flame speed S L and the quenching distance d q .
Flashback can also occur because of turbulent flame propagation in the core flow. Combustion instabilities have a very considerable effect on system dynamics and can cause flashback due to non-linear interaction of pressure fluctuations, hence periodic heat release and non linear flame propagation [15]. Finally flashback in swirl burners can be caused by a phenomena termed combustion induced vortex breakdown (CIVB) due to rapid expansion at the burner exit creating a recirculation zone which acts as a flame holder: the breakdown of this structure can occur due to flow perturbations and chemical reaction effects causing the CRZ and hence flame to propagate upstream into the premixing zone [16,17].

Experimental setup
The generic swirl burner was used to examine flame stability limits at atmospheric conditions (1 bar, 293 K). The was designed and assembled at Cardiff University's Gas Turbine Research Centre (GTRC). A single tangential inlet feeds an outer plenum chamber which uniformly distributes premixed air/fuel to the inserts, eventually into the burner body. A central fuel injector extended through the whole body of plenum and the insert burner. Principally, the fuel injector is used to produce both non-premixed and partially premixed flames; its position is shown in Figs. 1 and 2. This simulates many industrial applications where liquid fuels are sprayed through a central fuel injector.
Three swirl numbers have been used in the experiments, with the only change in the system being in the exhaust insert with tangential inlets which force flow into the swirl chamber, then exhaust. Three inserts are used with different swirl numbers, achieved by changing the number, length and width of the tangential inlets. The three swirl burners have swirl numbers of: S I = 1.47, S II = 1.04, S III = 0.8. Based on other work [9,21] an exhaust nozzle extension 0.5De long was added to the exhaust of two of the inserts. The fuel injector was left in the same position Swirl insert III is very similar to II the only differences lying in the width of the tangential inlets, 5 as opposed to 4 mm (nine inlets used). Swirl insert I only has four inlets, but operated at a significantly higher swirl number of 1.47, Fig. 2.
Coriolis flow metres have been used simultaneously to measure the mass flow rate of both fuel and air separately.

Results and discussion
Three swirl burners plus five different fuels has been used to obtain results, these are summarised in Tables 1 and 2: Typically the pressure loss coefficient at S II = 1.04 is nearly half that at S I = 1.47 and again is about 20% lower again at S III = 0.8. Lower pressure drop is a major advantage to designers and operators of gas turbines and other large burners and thus there is a drive to use lower swirl numbers, providing the flame stability advantages of the CRZ are not lost. coke oven gas has been used as a representative process industry fuel gas, which is widely   available at steelworks and has the potential to be widely used in power generation in process industry, providing appropriate efficient reliable technology can be developed to utilise it. The system has been tested on a wide range of fuel blends as shown below, Table 2. Up to 15 combinations of swirl burner and fuel gases have been used to investigate their effects on the flashback and blowoff characteristics. Fuel characteristics are interesting as they show similar lower heating values and adiabatic flame temperatures. The exception is pure hydrogen with much higher lower heating value, but adiabatic flame temperature about 100 K higher than coke oven gas. Three families of flashback curves are shown in Fig. 3 below, one for a swirl number of S I = 1.47, Fig. 3a, the other at a swirl number S II = 1.04, Fig. 3b and 3c for S III = 0.8.
Associated flame photographs at conditions just before flashback for pure methane are shown in Fig. 4a (S I = 1.46) and Fig. 4b (S II = 1.04).
The comparison is extremely interesting whilst other analysis has revealed two different flashback mechanisms for the different swirl numbers [10,[18][19][20][21]. With S I = 1.47 the central recirculation zone (CRZ) extends over the central fuel injector to the base plate for all fuels, with an associated flame front on the CRZ boundary. This is illustrated in Fig. 2 (and does not happen with S II = 1.04 and S III = 0.8). Flashback occurs when the radial velocity in the swirl level drops to such a level that the near radial flame front can flashback to the inlets and often into the plenum chamber [10]. Conversely with S II = 1.04 and S II = 0.8 flashback occurs by a different mechanism via flashback in the outer wall boundary layer of the exhaust nozzle, then being controlled by the critical boundary velocity gradient [21] as defined by Lewis and von Elbe [14]. This can be readily derived from geometrical and simple flow considerations and enables comparison with the large quantities of data available in past literature as summarised in [14]. Other work using CFD analysis of the boundary layer region close to flashback has shown that under the turbulent flow conditions of the swirl burner, critical boundary velocity gradients are an order of magnitude higher than those predicted by the Lewis and von Elbe formula [21]. In terms of flashback limits for methane and methane containing up to 30% hydrogen a value of S II = 1.04 and S III = 0.8 produces flashback which occurs at a mass flow (and hence velocity levels) up to 1/3 of those found for S I = 1.47 for a wide range of equivalence ratios. However with coke oven gas (COG) different effects start to appear as the hydrogen content of the fuel increase beyond 50%. For Swirl Numbers of 0.8 and 1.04 flashback performance is better than S = 1.47 for values of equivalence ratio up to 0.6 to 0.65 and mass flows of 7 g/s. Beyond this point for equivalence ratios > 0.65 and <1.2 a swirl number of 1.47 is better by up to 50%. However for LP combustors the aim is to operate around an equivalence ratio of 0. 7 or less and thus this is not a disadvantage. Comparison of the three Swirl Number cases, Fig. 3, shows that there is a significant change in flashback behaviour moving between a fuel with 30% hydrogen content to one with 65% hydrogen content as with COG. Moving onto the pure hydrogen results similar trends were evident, although the range of equivalence ratios tested was restricted to being below 0.5 and above 2 due to the very large hydrogen and air flow rates required. Here the higher mass flow, hence velocity levels, associated with hydrogen flashback, produce higher levels of turbulent kinetic energy, thus augmenting the turbulent flame speed and thus worsen the hydrogen flashback limits beyond that expected from considerations of laminar flame speed data [14,21].
More detailed inspection of the results for S II = 1.04 and S III = 0.8, showed generally both swirlers have very similar characteristics with differences being within experimental limits. S III = 0.8 is preferred as it gives lower pressure drop.
Another interesting result was that the peaks of the flashback curves tended to occur at weak equivalence ratios as opposed to the expected just on the rich side of stoichiometric [14]. This effect is thought to be due to changes in the recirculation zone occurring as the equivalence ratio approaches 1. This is also illustrated by Fig. 5 where all the methane data has been plotted as a function of critical boundary layer gradient at flashback, G f ; also included is laminar data on natural gas. The swirl burners at S II = 1.04 and S III = 0.8 are flashing back at lower values of G f than the laminar results (albeit at a higher pressure drop), whilst for S I = 1.47 values of G f are significantly higher.
Overall S III = 0.8 gives the best flashback limits for methane based fuels with hydrogen content up to 30% and for hydrogen based fuels with hydrogen content P 65% for equivalence ratios 6 0.65. However for fuels with hydrogen content in the range 30% 6 H 2 content 6 65% a more complex picture emerges. The Critical Boundary Velocity Gradient for flashback is higher at lower swirl numbers and equivalence ratios 1 when compared to S I = 1.47. Separate tests on blow off limits show that the Swirl Number S = 0.8 produces the best results.   A gas turbine, required to be dual fuelled, with given compressor and turbine system has air mass flow rates at given thermal inputs which vary little as the fuel mass flow is relatively small and the exhaust gas composition, hence enthalpy, is still dominated by the 80% nitrogen content from the air. To produce this thermal input different quantities of fuel and thus equivalence ratio are needed for different fuels such as natural gas, coke oven gas and especially pure hydrogen. When dual fuelling/changeover is needed ideally the operational range of the system between flashback and blow off for two different fuels (such as hydrogen and natural gas) should be such that there is sufficient overlap between the blow off and flashback limits to enable easy fuel change over. Because of the different stoichiometry and heating value, hydrogen containing fuels will always have to be operated at weaker equivalence ratios than natural gas fired systems, typically 78% of the natural gas equivalence ratio for pure hydrogen. This infers that the overlap region between the flashback limit and blow off limit of given fuels is crucial in determining whether or not the system can be dual fuelled. Table 2 indicates that because of similar adiabatic flame temperature and lower heating values fuel gases containing up to 65% hydrogen (as with coke oven gas) with a base fuel of natural gas can be best accommodated in existing or somewhat modified combustion systems.

Conclusion
This paper has discussed the flashback limits of three different swirl burners and shown that considerable differences exist. Preference is given to the system with low swirl number as it gives lowest pressure drop. The behaviour of methane based fuels with hydrogen content up to 30% has been shown to follow that of methane as the hydrogen content is increased. However coke oven gas shows distinctly different behavioural patterns, as does pure hydrogen which needs to be investigated further.