The effects of partial premixing on a reacting jet-in-crossflow is investigated in a five atmosphere axially staged combustor at stationary gas turbine relevant conditions. The facility consists of a dump style headend burner that provides a crossflow with a quasi-uniform velocity and temperature profile to the axial stage to isolate the effects of the jet-in-crossflow. The headend burner is run with methane and air at a lean equivalence ratio to match industry emission standards. For this work, the total air to the headend and axial stage is kept constant, and fuel is split between the headend and axial stage to represent different gas turbine loading conditions. For the cases analyzed, the fuel split to the axial stage went up to 25%. The axial stage consists of an optically accessible test section with a coaxial injector that provides variability to how long the methane and air can mix before entering the facility. Three different premixed levels are studied: fully premixed, nonpremixed, and partially premixed. The flow-field characteristics of the reacting jet-in-crossflow are analyzed using particle image velocimetry (PIV), and flame behavior is quantified by employing CH* chemiluminescence. NO measurements are made at the exit of the facility using a Horiba emissions analyzer. Two different flames are observed: flames that burn in the leeward recirculation region and flames that burn at the core of the jet.
The design and operation of power generating stationary gas turbines are driven by strict emissions regulations. In recent years, dry low NOx combustors have been able to achieve lower NOx output for the same turbine inlet temperatures by running at leaner equivalence ratios, replacing diffusion pilots with premixed pilots, using separate combustion systems, and controlling dilution air at part load [1–5]. Axially staged combustion systems have further improved the emissions output of these turbines over a more extensive range of operating loads. Axially staged combustion is effective in decreasing NOx emissions because of the reduced residence time of the fuel burned in the axial stage. However, reducing CO and NO formation are at odds with each other; as residence time is increased, hydrocarbons burn more completely mitigating CO formation. However, this causes high temperatures to be sustained for an extended period, which promotes thermal NO formation. To meet emissions requirements over a broader range of operating loads, gas turbines modulate the fuel/air split to the axial stage based on the turbine loading condition. At full power, fuel is bypassed to the axial stage to reduce residence time effectively reducing thermal NO formation. At part load, the main premixed stage is burned hotter and the axial stage is used as diluent air to improve CO emissions. This gives greater flexibility to operate the gas turbine over a broader range of operating loads.
Karim et al.  investigated the effects of axial stage fuel split on NOx and CO emissions in GEs 7HA class heavy duty gas turbine engine. At full load operation, as more fuel was bypassed to the axial stage, NOx decreased until an optimal point was reached and then began increasing again. This is attributed to the increase in axial firing temperature overpowering the decrease in NOx in the first stage (headend) beyond the optimum point. At part load, CO is the limiting factor for emissions compliance. At part-load conditions, axial fuel staging provided decreased exit temperature for a given CO level. At these conditions, the axial fuel stage was run with air only to allow higher firing temperatures in the headend to reduce CO. Additionally, the impacts of axial staging on combustion dynamics were examined. Over a wide range of fuel splits, axial staging had minimal effects on combustion dynamic amplitudes.
Recent studies have examined the relationship between emissions and flame behavior [7–9]. Flame behavior is crucial for combustor design as it is undesirable to have the flame burn at the walls or downstream in the turbine. Sirignano et al.  showed that rich methane-air jets initiate on the leeward side with a highly lifted flame on the windward side in an atmospheric combustor. As ΔT from the axial stage was increased, NOx was observed to increase monotonically. For a fixed increase in equivalence ratio across the axial stage, it was observed that rich jets produce less NOx at lower momentum flux ratios. Lamont et al.  found staging benefits for jet equivalence ratios between 0.41 and 1.12 and momentum flux ratios less than four in a low swirl crossflow in a 5.5 bar axially staged combustion facility. For rich jets with equivalence ratios ranging from 4 to 20, larger increases in outlet temperatures were observed, but NOx emissions also increased. Using CH* chemiluminescence, jet centerlines were obtained and compared to the nonreacting Holdeman correlation. For equivalence ratios near one, the centerlines did not agree with the Holdeman correlation, likely due to heat release causing dilatation. For rich jets, the Holdeman correlation adequately predicted the jet centerline trajectory.
Flame liftoff and jet centerline trajectory have been studied extensively by Wagner et al. [10,11]. For an ethylene-air reacting jet-in-crossflow, the leeward flame branch was observed to be more consistent than the windward flame branch. The windward branch showed a significant fluctuation in liftoff height, varying from anchored to lifted to completely blown out. Additionally, nonreacting jets-in-crossflows were compared to the reacting cases at the same momentum flux ratio. The reacting cases penetrated further into the crossflow than the nonreacting cases for low momentum flux ratios. As the momentum flux ratio was increased, the reacting and nonreacting jet trajectories began to converge. At a certain momentum flux ratio (J = 22.7), the trajectories collapsed onto each other.
Emissions of partially premixed flames have been widely studied [12–16]. Partially premixed flames are defined as flames where additional oxidizer is required to mix with a fuel flow to achieve complete combustion. Gore  found for laminar methane/air flames an optimal level of premixing where NOx emissions are at its minimum. Lyle et al.  reported no significant change in NOx emissions from fully nonpremixed jets (equivalence ratios greater than 5). Reduced NOx levels were observed for equivalence ratios between 1.5 and 5, where the lowest levels were observed at an equivalence ratio of 1.5. Kolb et al.  reported for lean reacting jets in crossflows, different timescales influence liftoff height including the Kolmogorov time scale, the chemical time scale, and the ignition delay timescale. It was found that predictions using the laminar flame speed did not accurately predict liftoff height.
In axially staged gas turbine combustors, the fuel and air in the axial stage are mixed as close to the combustor as possible to avoid premature combustion due to high temperatures generated by the compressor. This results in a partially premixed flame in the axial stage even if the jet equivalence ratio is near one. This configuration is fundamentally different from a rich jet in crossflow. A rich jet in vitiated crossflow is obtaining its oxygen from an oxygen-depleted source (the vitiated crossflow), whereas a partially premixed jet has an unburnt air source to mix with. The present research explores the effects of this partial premixing on flame liftoff height and flame structure and relates them to the flow-field dynamics. There is a desire to understand how the aforementioned timescales change with partial premixing in the axial stage at power generating gas turbine relevant conditions. The present findings will aid in predicting flame behavior based on jet conditions to improve combustor and injector design.
2 Experimental Methods
2.1 Experimental Facility.
The experimental facility is shown in Fig. 1. The facility consists of a main burner (headend) which provides vitiated crossflow for the axial stage. The headend burner consists of a 1.5″ main burner pipe which issues a methane-air mixture into a 2.5″ pipe creating a recirculation zone to stabilize the flame. Bypass air with 3 μm diameter aluminum oxide (Al2O3) particles for PIV is injected coaxially around the main burner pipe to lower the overall headend equivalence ratio and provide seeded air for the axial stage. The main burner pipe is run at lean conditions with its equivalence ratio ranging from 0.65 to 0.85; with the bypass air, the total headend equivalence ratio ranges from 0.58 to 0.73. All tests run in this research have a total air and fuel mass flow rate of the 0.73 equivalence ratio with different percent fuel splits to the axial stage to represent a gas turbine operating at baseload. As airflow to the axial stage is increased, the jet becomes highly unstable. Because of this, the airflow to the axial stage is kept constant for each condition.
Air is supplied by a set of high pressure compressed air tanks and is regulated using a Tescom (St. Louis, MO) dome pressure regulator. A second pressure regulator is used to fine-tune the air to the desired pressure for the USA Industries (Houston, TX) restriction orifice union, where the pressure measurement obtained from a Dwyer (Michigan City, IN) pressure transducer is used to calculate the mass flow rate. Similar networking is used to meter the methane to the facility. However, O'Keefe (Monroe, CT) precision orifices are used in place of the restriction orifice unions.
The headend burner has been tuned to produce temperature and NOx outputs that would be expected in an actual gas turbine. Center point temperature measurements are made at the inlet of the test section with an exposed bead b-type thermocouple, and a correction factor is applied to account for radiation heat losses. A Horiba (Kyoto, Japan) Mexa-540 L gas analyzer is used to obtain dry NO measurements at the test section inlet, and a 15% O2 correction factor is utilized. Figure 2 shows temperature and NOx curves with just the headend burner running to characterize the inlet boundary conditions for the axial stage. Also presented in Fig. 2 is the velocity profile obtained by the PIV at the test section inlet. Depending on the headend burning condition, the inlet velocity to the test section ranges from 65 to 80 m/s. The novelty of the facility is its ability to provide these uniform boundary conditions at 5 atm with the capability of running as high as 10 atm. This is achieved by leveraging a 26″ long mixing section after the headend burner that gives the flame ample time to fully burn and mix with the bypass air; a flow straightener is placed 6″ upstream of the test section to promote additional mixing.
The test section with the coaxial wall flush injector is presented in Fig. 3. Air seeded with the same particles as the main stage is supplied through the tee on the left side, while fuel is supplied through a bore-through Swagelok from the top of the tee. The tube inside the fitting can be translated up and down to change the distance the fuel and air mix (the variable y in the figure). For the partially premixed case y = 5″, and the nonpremixed case y = 0″ (the fuel tube is flush with the top wall). The bore-through Swagelok is removed from the top for the fully premixed case, and the top of the tee is capped; the fuel and air are mixed sufficiently upstream (L/D > 100). The top plate where this injector is attached is interchangeable to allow for different injector geometries to be studied. This work analyzes a 0.5-in. wall flush injector. The test section has three quartz glass windows, one on the bottom and two on the sides, for laser and imaging diagnostics. At the end of the test section is a choke plate that is leveraged to choke the flow to the desired pressure (in this case, five atmosphere). At the exit of the facility is an emissions probe that is immediately brought to an ice bath to freeze NOx production and condense out the water.
2.2 Experimental Diagnostics.
The total run time of the facility is 4 s for each test. CH* chemiluminescence is used to capture the structure of the flame. Imaging is taken using a high-speed Photron (Tokyo, Japan) SA1.1 Fastcam with a narrow band 430 ± 2 nm filter at 125 frames per second with a shutter speed of 1/1000 s. The images were recorded at a resolution of 768 × 768 pixels, which corresponds to 122 pixels per inch. A constant threshold is used to determine the flame location for each case.
Particle image velocimetry is used to obtain velocity fields. A 532 nm dual-head Evergreen (Durham, CT) laser is fired with a time separation of 20 μs between each laser. The beam is focused through a 1000 mm spherical focusing lens and then passed through a -25.4 mm cylindrical lens to generate the laser sheet. The facility is located just under 1000 mm from the focusing lens to obtain maximum power in the test section. An Andor (Abingdon, UK) camera with a narrowband 532 + 10 nm filter captures the seed particle movements. Imaging is acquired at 12.5 frames per second using the frame straddling method at a resolution of 2048 × 2048 pixels which corresponds to 630 pixels per inch. Aluminum oxide (Al2O3) particles with 3 μm diameters are used to seed both the crossflow and the axial stage airflow.
In the headend section, the seed is injected through the bypass air and mixes with the flame through a perforated screen in a 26″ mixing section. The jet is seeded by rerouting a percentage of air to a separate seeder and is then mixed with the unseeded air upstream of the injector. The percentage split to the seeder varies based on the conditions. A BNC model 575 pulse delay generator is used to synchronize the cameras and lasers. The data is processed in PIVlab 2.00; a contrast limited adaptive histogram equalization method (CLAHE) is leveraged to increase the contrast of the image to improve seed visibility. A four-step multipass method is employed with the largest pass of 198x198 pixels and the smallest pass of 32x32 pixels with a 50% overlap at each step.
As previously mentioned, the total fuel and air run through the facility is kept constant at an overall equivalence ratio of 0.73, which corresponds to a total firing temperature of 1650 °C. The air to the headend and axial stage is fixed, and the fuel is modulated between the two stages. Time-averaged CH* contours of 15% and 25% fuel splits are presented in Fig. 4. The 15% and 25% fuel splits correspond to jet equivalence ratios of 1.07 and 1.78, respectively. For the 15% fuel split (the leaner jet), the crossflow equivalence ratio is 0.69 which produces a temperature of 1475 °C; for the higher fuel split, the crossflow equivalence ratio is 0.61, which provides a temperature of 1400 °C. For the 15% fuel split, the crossflow Re = 97,710 and the jet Re = 90,519. The 25% fuel split corresponds to a crossflow Re = 97,649 and a jet Re = 92,276. For both cases, because the air to the axial stage is kept constant, the momentum flux ratio is the same J = 5. For each premixed level, the Φ = 1.78 jets feature a flame with its core downstream of the viewing window. The fully premixed and partially premixed Φ = 1.07 jets feature a flame with its core inside of the viewing window, while the nonpremixed case burns downstream mostly outside the viewing window. As expected, the nonpremixed jets exhibit a highly lifted flame, while the fully and partially premixed flames are only slightly lifted. Time-averaged flame liftoff height for each case is determined by marking the initial point on the CH* image where the flame threshold is met, using the PIV to trace a streamline to that point, and determining the distance.
Figure 5 shows the liftoff height for the two fuel splits, each with three different premixed levels. For both jet equivalence ratios, the partially premixed and fully premixed cases did not vary too much in liftoff height, both lifting off about two jet diameters. This indicates either 5″ is a sufficient mixing length for the given conditions or the timescale associated with the heat transfer to the jet to reach the ignition temperature is greater than the timescale associated with the jet mixing for the partially premixed case. Both nonpremixed cases are lifted over four jet diameters. For the higher fuel split each flame is lifted more than its leaner jet counterpart. For the higher fuel split, the crossflow equivalence ratio is less than the lower splits resulting in a colder headend flame. Previous work has shown a strong correlation between liftoff height and crossflow temperature. This work shows for different premixed levels this trend still holds. Liftoff height also shows a strong dependence on jet composition . Methane-air flames tend to liftoff higher than flames with hydrogen mixed in. Also presented in the figure are liftoff heights for rich methane-air jets studied by Sirignano et al. . In this work, the nonpremixed flames at equivalence ratios of 1.07 and 1.78 lift off as high as a rich jet with equivalence ratios between 3 and 5. The above discussion gives insight into the time-averaged liftoff. This is mainly dominated by the leeward branch which is inherently more stable than its windward counterpart; to investigate the fluctuations of the windward branch, a root-mean-square of the windward liftoff is calculated for the fully premixed Φ = 1.07 case. The root-mean-square for this case is 0.86 jet diameters. This significant fluctuation in windward liftoff is consistent with what is seen in previous investigations.
Figure 6 shows the time-averaged vorticity field with the CH* flame trace overlayed onto it. Two distinct vorticity features appear in the time-averaged data: the shear layer on the windward side of the jet, and the shear layer on the leeward side of the jet. For the fully premixed and partially premixed cases for both jet equivalence ratios, the flames initiate on the leeward side of the jet. However, for the jet Φ = 1.07, the flame spreads to the edge of both shear layers while the flame of the jet Φ = 1.78 resides mainly in the leeward shear layer. Rich jets are known to stabilize on the leeward side . This is likely due to the low-velocity region created by the recirculation zone of the jet. In this case, the jet is 0.5 in which is considerably larger than previous investigations [7–9]. As a result, the recirculation zone is larger, which creates a favorable environment for a flame to stabilize. To further investigate this trend, Fig. 7 demonstrates the probability density function of the vorticity inside the flame for the two different jet equivalence ratios. The richer jet that stabilizes in the leeward side is heavily skewed to the left indicating a strong presence of negative vorticity within the flame (consistent with what is seen in Fig. 6). The leaner jet, which stabilizes at its core, exhibits a more even distribution being only slightly skewed to the right. This indicates the presence of both negative and positive vorticity within the flame as depicted in Fig. 6.
Although the flame for the leaner case initiates on the leeward side of the jet, it spreads to the jet core, and its edges burn in the shear layer on both the leeward and windward sides. There are two fundamental differences for the leaner case flame which explain this phenomenon: the leaner jet is at a more reactive equivalence ratio (Φ = 1.07 versus Φ = 1.78), and the leaner jet corresponds to a lower fuel split, so the crossflow temperature is hotter (1400 °C versus 1475 °C). Contrary to the fully premixed and partially premixed cases, both nonpremixed cases ignite in the core and spread to the edge of both the leeward and windward shear layers downstream. This insinuates the timescale associated with the mixing of the fuel and air supersedes the timescale associated with heating the jet to the ignition temperature. In these cases, the jet reaches its ignition temperature before the fuel-air mixture reaches an ignitable mixture fraction. This leads to a jet that is hot enough to burn at its core when an ignitable mixture fraction is reached. Both these cases ignite close to the end of the viewing window with most of the flame downstream of the test section.
Flame ignition for reacting jets in crossflows is generally broken into two regimes, auto-ignition and flame propagation . For auto-ignition to occur, specific conditions must be met, including adequate temperature (sufficient heat transfer from the crossflow to the jet), specific radicals, and sufficient time at these conditions. Turbulent mixing between the jet and the crossflow accelerates the process to meet these requirements. Once these conditions are met, the spot with the most reactive mixture fraction (MRMF) ignites. Numerical studies have attempted to model these MRMF for various conditions [20–23]. These MRMF spots are generally very lean. At the acquisition speeds run in this work, it is not possible to determine the ignition regime of the flames. However, the richer jets stabilizing in the leeward shear layer suggest the mode of flame ignition could be flame propagation, while the leaner jets burning in the core could indicate auto-ignition events. The lower fuel split creates a more conducive environment for auto-ignition events compared to the higher fuel split. The hotter crossflow and leaner jet equivalence ratio could explain why the lower fuel split burns at the core as opposed to the higher fuel split that stabilizes in the leeward shear layer. Further investigation is required to definitively conclude these hypotheses with high-speed planar laser-induced fluorescence preferred to capture the flame structure over line-of-sight chemiluminescence.
The PIV jet centerline trajectory for the fully premixed jet, the nonpremixed jet, and the Lefebvre nonreacting centerline correlation  are presented in Fig. 8. The penetration depth for each case is determined by following the jet's centerline streamline, and determining where the v-component of velocity merges with the crossflow. Each case represents the same momentum flux, J = 5, but have different flame structures/equivalence ratios. For the fully premixed 15% fuel split, where the flame burns in the core of the jet, the jet significantly over penetrates the nonreacting maximum penetration depth. The leeward stabilized fully premixed 25% fuel split over penetrates compared to the nonreacting penetration correlation, but not by a noticeable amount. Along with this, each jet penetrates further into the flow before its trajectory begins to turn compared to the nonreacting centerline correlation. Although the lifted flames do not burn until further downstream, they are being heated by the high-temperature crossflow which results in a velocity increase. Knowing the penetration depth of the flame in the flow-field is of particular interest because if the flame over penetrates, it could potentially burn at the combustor wall causing unwanted damage. Nonreacting jet trajectory and penetration correlations are currently used in industry to model reacting flow-fields. More advanced models are required to accurately predict the trajectory of reacting jet in crossflows.
3 Conclusion and Future Work
The effects of premixing and fuel split on flame and flow-field characteristics were analyzed in a high pressure axially staged combustion facility. CH* chemiluminescence was leveraged to obtain the structural features of the flame, while PIV was utilized to quantify flow-field attributes. Liftoff showed a strong correlation to how well the fuel and air were premixed coming into the facility, along with the jet equivalence ratio; this is in agreement with previous investigations. The nonpremixed jets lifted off as high as jets with equivalence ratios as high as five, despite having equivalence ratios close to one, suggesting multiple timescales affect the ignition delay time of the jet. The time-averaged vorticity field with the CH* chemiluminescence overlaid gave insight into where and how these flames are stabilized. The richer jet burns in the low-velocity leeward shear layer, while the leaner jet ignites in the low-velocity leeward region but the flame spreads to its core. Finally, the jet centerlines over penetrated the nonreacting correlations due to the presence of heat release, which causes dilatation; the effects of this were more pronounced for the jet with the equivalence ratio closer to one.
The data presented in this paper is characterized to be applicable to dry low NOx power generating gas turbine combustors. For these combustors, it is crucial to have a lifted flame to ensure no burning occurs at the walls of the combustor. For each case analyzed, the observed flames were lifted at least one jet diameter. This study observes rich jets (jets with equivalence ratios greater than unity); the jet with an equivalence ratio closer to one ignited in the shear layer and spread to the jet core burning mostly within the viewing window. The richer jet ignited in the shear layer, and the flame remained mainly inside the shear layer burning primarily downstream of the viewing window. While a lifted flame is desired to avoid combustor damage, it is undesirable to have a flame that burns slowly and inefficiently. This can cause issues if the flame burns downstream in the turbine. Along with this, slow-burning flames lead to increased thermal NO formation. The richer jet in this study exhibits characteristics that suggest this kind of burning.
Future investigations include sweeping a more extensive range of jet equivalence ratios, including lean jets and jets richer than the ones analyzed in this study. Additionally, emissions measurements with the axial stage running will provide further insight into how the results found in this work relate to emissions output, and where the optimal fuel split is for NOx formation. Finally, high-speed OH planar laser-induced fluorescence and high-speed PIV will provide further understanding into the ignition mechanisms of these flames and their associated flow-fields at various conditions.
The authors would like to thank Dr. Carlos Velez and Dr. Keith McManus of GE Global Research for their provided input and guidance. This work is supported by the Department of Energy (grant number DE-FE0031227, Dr. Seth Lawson COR. Any opinions, findings, and conclusions or recommendations expressed in the material are those of the authors and do not necessarily reflect the views of the Department of Energy.
U.S. Department of Energy (Grant No. FE0031227; Funder ID: 10.13039/100000015).