The Effect of Pressure on NO_x Entitlement and Reaction Timescales in a Premixed Axial Jet-In-Crossflow

The present work investigates the influence of pressure on the reaction timescales and NO_x emission of a premixed jet in vitiated crossflow. The present study attempts to fill a gap about axial-staged combustor knowledge since a quantitative statement about the influence of pressure on NO_x emission has not been made in the open literature. The outcomes could be of high interest to further understand how applicable the data of flames at feasible laboratory pressure levels between 1 atm and 5 atm would be for burners at pressure levels 20–30 atm applied industrially. Patents in the field of axial-staged gas turbine combustion are owned by the OEMs Siemens [1,2] and GE [3,4].

On the experimental side, recent studies by Elkady et al. [5] investigated a perfectly premixed research burner operated at high-pressure levels between 8 atm and 16 atm. It was concluded that a pressure dependency of NO_x emission X_NO ∼ p^const can be formulated, but the constant would be temperature dependent. Higher pressure was shown to reduce both main stage burner NO_x and CO emission levels experimentally [5] with a pressure dependency in the range of p^0.5 for the main stage burner. Research by Gollahalli and Pardiwalla [6] focused on the influence of jet shape on the emission profiles of turbulent propane jet flames in crossflow. Experiments were conducted in a wind tunnel with an optically accessible test section attached. The upper and lower operating limits were screened by variation of the jet velocity. Lower CO emission and higher NO emission levels were the result of using elliptic jets relative to circular shaped jets [6]. A novel approach to obtain more reliable experimental NO_x emission monitoring was suggested by Al-Malak et al. [7], describing a soft emission analyzer that utilizes an artificial neural network. Blowout limits of a methane jet diffusion flame surrounded by smaller fuel-flex jet pilot flames and air co-flow were studied experimentally by Kibrya and Karim [8]. The blowout limit was extended by increasing the pilot jet flow, allowing enhanced entrainment and higher thermal contribution of the side jets [8]. Flame hysteresis effects of methane jet flames with variable air co-flow were studied experimentally by Moore et al. [9] to determine characteristics of the instantaneous turbulent flame liftoff relative to the fuel velocity. Critical operating parameters of lean partially premixed methane–air turbulent combustion were investigated experimentally by Askari et al. [10] in a constant volume combustion chamber and visualized with a Schlieren shadowgraph. The effect of hydrogen enrichment, equivalence ratio, diluent addition, variation of injection pressure, chamber pressure, temperature, and turbulence intensity (TI) were investigated. The addition of hydrogen was reported to be the critical parameter, allowing to improve the lean-burn limit of CH₄/air, improving peak p and T, and decreasing the duration of combustion cycles [10]. The utilization of process synergies and the integration of gasification into a combined cycle heavy-duty gas turbine system were studied experimentally and numerically by Liu et al. [11]. Required adaptations of conventional combustors to allow operation with syngas include the combustor head and nozzle arrangements. Numerical results were verified with experimental data, and a design was proposed for the mid- and high-pressure range.

Recent modeling efforts include a numerical large eddy simulation (LES) investigation with star-ccm+ by Elgammal and Amano [12] on the thermal uniformity and pressure drop of jet-in-crossflow (JIC) mixing in a central swirler with staggered holes and various designs were able to document the axial and radial profiles with increased fidelity. The effects of distributed methane combustion on the temperature profile and thermal NO_x emission profile were investigated experimentally and numerically by Khalil et al. [13]. The distribution profiles were adjusted with a swirl field and variation of the CH₄/air premix state under lean conditions. Outcomes improved the pattern factor and achieved CO and NO_x emission levels in the low ppm range [13]. In a subsequent study by Khalil and Gupta [14], the influence of premixed fuel/air dual injection in the swirl field was documented and found to further reduce emissions relative to the single injection pattern studied by Khalil et al. [13]. A non-premixed turbulent jet flame (Sandia Flame D) was simulated with an experimentally verified LES by Safari and Sheikhi [15] to determine process irreversibilities and amount of entropy generation. Further numerical investigations by Deng et al. [16] covered the impact of nozzle configuration, such as the number and pattern of nozzles, the use of premixed and non-premixed nozzles, and jet velocity by applying moderate or intense low-oxygen dilution (MILD) principles to an F class gas turbine combustor. Simulation at industrially relevant condition p = 16.3 bar, air temperature of 723 K, and a lean equivalence ratio of 0.625 suggested a number of 12 nozzles and jet velocity 110 m/s to be the optimal [16]. Simulations of an industrial gas turbine combustor were performed by Emami et al. [17] with gaseous CH₄ fuel under variation of the air flow distribution, recirculation, and dilution characteristics to optimize for CO and NO_x emission. Thermal NO_x and prompt NO_x formation mechanisms were considered, and the results emphasized the strong dependency of thermal NO_x levels on the local firing temperature and amount of air dilution [17].

τ_chem depends on the local CO₂ production rate, CO₂ molecular weight, local CO₂ molar concentration, and a very small number ɛ to avoid definition errors (1 × 10⁻¹²). The Damköhler number is used in this study to characterize the variation of the flame type depending on the system pressure. τ_mix is a moderate function of pressure, meanwhile τ_chem was shown to be strongly pressure dependent. A chemical activity profile for a one-dimensional (1D) premixed laminar flame combustor and 250 dynamic grid points is shown in Fig. 1. The significant differences of maximum heat release levels as well as streamwise length of heat release indicate the necessity for a three-dimensional (3D) analysis for the relatively complex jet-in-crossflow field. The second objective of this paper was hereof defined to investigate the transition between shear layer burning at low pressure and a receded jet core flame at elevated pressure level using a timescale analysis.

The experimental facility is displayed in Fig. 2. The facility consists of a main stage burner where a burner pipe houses a lean premixed methane–air mixture. A perforated screen is located between the main and axial stage to aid flow uniformity. The test section contains three optically accessible windows for imaging diagnostics, and an interchangeable top plate used to introduce fully premixed mixture to the facility in the second stage. The axial injector has an exit diameter of 12.7 mm. At the downstream end of the facility, a choke plate with a diameter of 38.1 mm is used to choke the flow to the desired pressure level. Further details about the experimental setup and procedure could be found in Ref. [26].

At the exit of the choke pipe, a probe is placed in the center to capture exhaust gases for emissions measurements. The probe is large enough to capture a significant amount of gas to obtain a global average. Tests were done where the probe was traversed and the spatial location at the choke point did not have a significant effect on the measurements, confirming the obtained values accurately represent a global outlet average. The gases are carried through piping into a large ice bath to freeze any further reactions from occurring and condense out the water. A Horiba Mexa 584-L emissions analyzer was used to obtain nitrogen oxide levels. Unless mentioned otherwise, emission levels in this paper are dry and corrected to 15% oxygen. In the computational fluid dynamics (CFD), the main stage boundary was defined using the measured data. The dilution average was taken by evaluating mass-weighted surface average function one diameter after the jet penetrates. Outlet probe in the CFD was determined as a mass-weighted surface average at similar position six jet diameters downstream of the choke plate. Referenced probe positions are shown in Fig. 3.

Line-of-sight CH* chemiluminescence data at 200 μm per pixel spatial resolution were taken using a Photron Fastcam SA1.1 at a rate of 125 frames per second (fps) and a shutter speed of 1/1000; a narrowband 430 ± 2 nm filter was utilized to obtain CH* intensity. Since the CH* measured in the experiment corresponds to the chemiluminescent emission from CH species, the measured CH* concentration follows the CH concentration in the flame and a direct comparison with simulated CH species was deemed valid [27]. To replicate the experimental line-of-sight technique, simulated CH data were integrated along the z-axis. A Δy/d = 0.05 was selected for the integration, weighting all z-data equally. Accumulated sum of z and Δy data was normalized against the maximal area sum. Images were recorded with a resolution of 768 × 768 and were processed in matlab to obtain time-averaged images considering one second worth of data for each testing condition. The flame boundaries are determined based on Otsu’s threshold method [28]. Particle image velocimetry (PIV) was performed using a dual-head 532 nm Evergreen laser with a time separation of 20 μs. The PIV data are processed in pivlab 2.00 using a contrast adaptive histogram equalization method (CLAHE). Figure 4 shows the overlay of planar turbulence intensity PIV data (blue plane) with simulated turbulence intensity (yellow-red 3D field), allowing an accurate representation of simulated data located at the z-symmetry plane.

The estimate is reasonable, given that steady-state conditions are not reached due to rig run times being below 5 s paired with a strong main stage burner temperature profile. In addition, CFD and experimental data do not suggest any significant flame–wall interaction to occur along the windows [31]. Further downstream, 304 steel is the major wall component in use with a higher k = 20 W/(m K) at 700 K and k = 16 W/(m K) at 300 K [32,33]. The jet walls were defined at constant T = 295 K. The dump region (2) in Fig. 5 is specified as a separate external flow zone. The plane surrounding the interface boundary was defined as an inlet at stagnating velocity and ambient condition. Boundaries downstream of that plane are defined as outlets at ambient pressure.

The hot gas composition entering from the main burner was determined with an equilibrium calculation, using a constant pressure/enthalpy reactor. Equilibrium mass fractions of nine species CO, CO₂, H₂, H₂O, HO₂, N₂, O, O₂, and OH were complemented with measured NO and NO₂ main stage burner levels. Results for the main stage species boundary condition at p = 5 atm are summarized in Table 1. A weak pressure dependency was found for the equilibrium at lower pressure, and main species remained near constant.

A coupled flow model with multi-component diffusion was required to account for the compressible flow field and the domain of choked flow. Material properties were determined with mass-weighted mixture and Sutherland’s law method [40]. The k–ω γ-Re_θ shear stress transport (SST) model was used and in relatively good agreement with the data. At the main stage outlet, a developing flow profile was determined for all pressure levels investigated. PIV velocity profiles are shown in Fig. 7. At 1 atm, the mass flows needed to be increased relative to the scaling law due to stability limits of the experimental high-pressure facility. Despite the increase of mass flow, the momentum flux ratio remained constant among the cases. Profiles at 3 atm and 5 atm were near-identical due to the linearly scaled condition.

Maximum uncertainty was calculated to be 13% from the PIV data located in the near-wall domain. The measured main stage PIV profiles are reasonable, given the short entry length of x/d_h = 10 from the main stage inlets. Profiles in Fig. 7 defined the CFD crossflow boundary condition, and the γ-Re_θ transition function was defined to allow up to 25 mm thickness of hydrodynamic boundary layer. An accurate description of the reacting jet trajectory was achieved and proven in previous work [36]. The crossflow TI profile was determined from the PIV to rise from 15% at the centerline to about 30% along the wall. Turbulence intensity at the jet was set to 6%, and turbulent length scales were set to L = 5 mm at the main stage boundary and L = 1 mm at the jet, respectively.

A grid convergence study was done by variation of the grid refinement in the critical jet and reactive domain between levels Δs = 0.25 mm (29 million cells), Δs = 0.5 mm (4.5 million cells), and Δs = 1.0 mm (0.7 million cells). Target was to show the influence of coarser mesh size and to determine possible tradeoffs that could be acceptable to mesh larger combustors for industry use. Figure 8 shows the CO₂ production rate plotted at iso-levels 1 kg/(m³ s), 10 kg/(m³ s), and 20 kg/(m³ s).

Whereas a sharp iso-levels contour at the upstream-most position was determined for the Δs = 0.25 mm grid (Fig. 8, red lines), some amount of numerical dissipation was recorded with the 0.5 mm grid (Fig. 8, blue crosses) and significant deviations were shown for the 1 mm grid (Fig. 8, green dots) in terms of the flame position and production levels. Details about the flame-counter rotating vortex pair (CVP) interaction are depicted less accurately upon transition to the coarser grids. Horizontal flame liftoff is strongly dependent on the mesh grid used and was shown to be 4% overpredicted for the 0.5 mm grid and 19% overpredicted with the 1 mm grid (relative to the 0.25 mm grid). Vertical flame liftoff Δy was relatively unaffected. Maximum CO₂ production rate is strongly grid dependent, transition from Δs = 0.25 mm to 0.5 mm reduced the maximum rate by 9%, and transition from 0.25 mm to 1 mm reduced the rate by 32%. End positions of significant CO₂ production were found to be 8% delayed for the 1 mm grid and 2% delayed for the 0.5 mm grid relative to the 0.25 mm grid. A CentOS 7 cluster was used with Intel Xeon cores, x86_64 architecture with 28–32 cores per node and 128–192 GB RAM. A Lustre parallel file system with 56 Gbit/s 4x FDR InfiniBand network fabric was used to transfer data. Power demand for a detailed chemistry simulation with 29 million cells was 20,000 CPU hours at dampened under-relaxation factor settings.

Mass flows were linearly scaled to operate at various pressure levels, keeping the fuel split, equivalence ratios, and flow profiles near constant. The influence of pressure reduction from 5 atm to 3 atm and 1 atm was investigated with a lean φ_main = 0.58 and a lean premixed φ_Jet = 0.75 axial jet. The boundary conditions used to simulate the investigated operating points are stated in Table 2.

A comparison of the simulated and experimental NO levels determined at the choke pipe outlet position (as described in Sec. 2.1) is summarized in Table 3.

Experimental pressure measurement was relatively accurate with a deviation of ±25 kPa at pressure and ±35 kPa at atmospheric condition. Accuracy of the experimental NO measurement was limited by the accuracy of the gas analyzer; increased positive emission uncertainty was stated for the experimental 1 atm condition since the combustion event may have been just near complete in the given facility length. Measured outlet unburned hydrocarbons (UHC) levels remained low (3 ppm).

The influence of pressure reduction from 5 atm to 1 atm on the reaction timescale and local NO levels was investigated in Figs. 10(A) and 10(B). To characterize the flame, CFD data for the molar concentration of CH were processed by evaluating line integrals along the z-direction at x/d = 5 (Fig. 10(C)). Uncertainty of CH* signal (Figs. 10(C) and 11(C)) ranged up to I/I_max = ±0.1 due to turbulent fluctuations of the flame and flow field. To obtain the correct result for the x/d plane (Fig. 10(C)), a Δy = 0.05 was selected and all z-values in the increment added up and normalized with respect to the sum at the maximum concentration level (compare with Sec. 2.2). Figures 10(A) and 10(B) (p = 5 atm, p = 1 atm) plot normalized CH and NO mole fraction data along y and z at a position x/d = 5 downstream of the jet. The 5 atm core flame shows NO production levels (blue lines) in spatial agreement with the CH levels (red line). Maximum NO level in the core is 23 ppm, and pronounced spread is found at the windward side due to proximity of the hot crossflow. At 1 atm, the near-flame domain (green line) separates two branches of NO occurrence (blue lines). Maximum NO level of 98 ppm can be found at the leeward side, indicating that convective transport at the CVP center is not enough to entirely inhibit NO levels accumulating in the domain between the flame front and the vortex center. Figure 10(C) shows the verification of z-integrated CH from the CFD with experimental CH* chemiluminescence data at the position x/d = 5, plotting the y/d axis against the normalized intensity level. Both CFD flames show CH levels in a domain between y/d = 0 and y/d = 4 and follow the measured data with deviations below 20%. Maximum z-accumulated CH intensity can be found at y/d = 2 for the shear layer flame (p = 1 atm), denoted by the green line in Fig. 10(C). At p = 5 atm, the red line (Fig. 10(C)) represents a higher maximum CH level for the core flame, found at y/d = 2.5. At x/d = 5 (Fig. 10(A)), the p = 5 atm data show a core flame burning close to the symmetry plane, exhibiting minor changes in the CH profile by the contribution of the z-data. The line integration method described above was particularly critical to match the p = 1 atm flame (Fig. 10(B)), considering shear layer burning with a contribution of simulated CH signals up to two jet diameters off the symmetry plane.

A similar analysis from a top view (at constant position y/d = 2) is depicted in Fig. 11. Compared were simulated CH and experimental CH* chemiluminescence data at p = 5 atm (Fig. 11(A)) and p = 1 atm (Fig. 11(B)) as well as the z-integrated CH signal for verification (Fig. 11(C)). At p = 5 atm, the CH tailing phenomenon shown in Fig. 11(C) can be tracked in the x,z-space: Fig. 11(A) shows decaying NO levels toward the outlet. This behavior suggests produced NO to be either transported to higher y/d positions by the CVP or to be further oxidized to NO₂ in the low temperature domain facing the bottom wall. The p = 1 atm data (Fig. 11(B)) validate statements made for Fig. 10. The top view of the two NO branches is shown to be separated by the shear layer flame front. NO levels trapped at the inside of the CVP reach peak levels of 67 ppm. A less steep slope with lower maximum intensity at x/d = 6 proves the slower reaction progress in the shear layer. CH levels were fully consumed before x/d = 7 (Fig. 11(C), light green), and this behavior was verified with the CFD (Fig. 11(C), dark green).

A comparison of the 5 atm and 1 atm total peak areas in Figs. 10 and 11 indicates that similar total CH amounts are present throughout the plane, and the z-data revealed that local distribution plays a significant role for the emissions results. Based on the data analyzed, pressure is seen to aid with lowering the production of NO_x emissions in the axial stage as a result of the occurring flame stabilization mechanism. Experimental data underline the significant influence of pressure on emission levels. At p = 5 atm, 5.0 ppm NO at 15% oxygen was measured after the main stage and 5.5 ppm at the axial stage outlet. In contrast, at atmospheric pressure, 2.5 ppm was determined after the main stage and 14 ppm at the axial outlet. The concept of axial staging is relatively ineffective at low pressure but significant emission saving potential was proven at lab-scale high-pressure level 5 atm.

To get a certain amount of insight to the turbulent scales, simulated and experimental turbulence intensities at the lee-side recirculation zone (x/d = 2.5) are compared in Fig. 12. Experimental PIV data were time-averaged and filtered to obtain relatively smooth turbulence intensity profiles with maximum errors of 20% (Fig. 12(a)). A significant pressure dependency of local turbulence intensities at the jet lee-side was determined. Maximum intensity levels at a position y/d = 1.7 were found to be 60% at 5 atm, 75% at 3 atm, and 115% at atmospheric condition. Beyond the windward trajectory edge, turbulence intensity shows a weak pressure dependency within levels of 20% and 25% (Fig. 12(a)), and these levels were found in the CFD as well. Figure 12(b) shows the simulated y,z field of turbulence intensity 50% iso-levels at the similar position x/d = 2.5. The CFD trend shows a more pronounced recirculation zone at lower pressure level. Since the PIV technique is well focused on the centerline by using a collimated laser sheet, the centerline axis (z = 0) of simulated 50% iso-level can be used for comparison with the profile found at a TI level of 50% in the PIV data. The evaluated position for this perpendicular comparison is highlighted with the green dashed lines in Figs. 12(a) and 12(b). The line profiles are relatively similar and suggest the RANS k–ω γ-Re_θ SST model with intensity and dissipation boundary conditions defined in Sec. 3.3 to give a good approximate description of the turbulent scales and their variation with system pressure.

The analysis of the present time scales at different pressure levels is described in this section. Averaged trends of the Damköhler number throughout the combustor are investigated in Fig. 13.

Mass-weighted y,z surface average of the Damköhler number was evaluated with an increment of x/d = 1. For the p = 5 atm condition, average was evaluated with an increment x/d = 0.5 to capture the steep rise of Da at the flame. At p = 5 atm, a distinct maximum was recorded with a certain amount of tailing toward the combustor outlet. At p = 3 atm, the similar trend was found but dampened due to the lower operating pressure. In contrast, a significant deviation is shown at atmospheric pressure. Combustion event is delayed, and maximum Damköhler number is reached after x/d = 10.

Due to the occurrence of local combustion event along the jet shear layer (particularly at low pressure), further analysis is required to quantitatively compare Damköhler numbers. Figure 14 shows the y- and z-directions at two distinct positions x/d = 5 and 10 downstream of the axial jet. At p = 5 atm, elevated local Damköhler number Da = 0.31 was found after x/d = 5 at the flame, located at the windward trajectory edge. A certain amount of jet core burning was recorded as shown by the relatively wide reactive domain. Combustion is near complete at x/d = 10, and production of post-flame NO remains as a weak signal. At p = 3 atm, maximum local Damköhler number Da = 0.095 is found at the windward edge. Reactions take place in a thinner domain at p = 3 atm (Fig. 14) that surrounds the jet more. Similarly, reaction is relatively complete at x/d = 10. Pressure reduction to p = 1 atm shows a significantly lower Da = 0.012 and distinct change in the reaction kinetics. Shear layer burning is concentrated in the CVP, and the flame may interact locally with the vortex center. At p = 1 atm, reaction progress is slow and significant reaction activity is present at x/d = 10. Applying the gained knowledge with the analysis in Secs. 4.1 and 4.2, the reason for high NO formation at low pressure is shown in Fig. 14. Due to the reduced bimolecular collision rate as a function of Da ∼ p², a longer (x/d direction) flame with more radial surface area (y/d, z/d) produces more radicals and favors elevated NO production. In contrast, at p = 5 atm, a relatively short, receded jet core flame is located at the windward trajectory side, allowing a significantly smaller flame surface and reduced NO levels, despite the locally higher temperature level.

This paper investigates lean premixed 12.7 mm axial jets in vitiated crossflow with detailed chemistry and 53 species considered. All operating points were validated against in-house experimental data and found to be in the finite-rate regime. Simulated exit nitrogen oxide levels were validated with experimental emission data, and a global emission trend for the pressure dependency of corrected nitrogen oxide emission was found to be X_NO,global = f(p^−0.55) for the investigated turbine inlet temperature level T = 1700 K. A substantial difference about the reaction kinetics and local nitrogen oxide production was pointed out for the pressure dependent transition between two flame types. At elevated pressure of 5 atm, jet core burning was determined at relatively high Damköhler number of 0.3. A receded core flame with a compact flame surface produced a moderate amount of NO emission along the jet core. High windward thermal NO emission was avoided due to the moderate temperature level used in this study. At atmospheric pressure, prolonged shear layer burning at low Da = 0.015 was determined to cause the elevated amount of nitrogen oxide emission. The physical reason was tracked with an analysis of the reaction time scale. The flame surrounded the entire shear layer and combustion required twice the axial length relative to combustion at pressure. Result was a large flame surface and high accumulation of NO at both sides of the shear layer. A Da ∼ p² dependency of the local Damköhler number was determined for the transition between shear layer burning and jet core burning. Chemical timescales were shown to match with experimental data by an analysis of the CH* chemiluminescence signal and simulated CH concentration in the flame. In contrast, a moderate pressure dependency was determined for the contribution of the turbulent timescales. Higher turbulence intensity along the lee-side recirculation zone of the jet at lower pressure was shown with an overlay of the PIV data and simulated turbulence intensities.

The authors (BS, TG, MO, KA, and SM) acknowledge support from the Department of Energy under Award Number DE-FE0031227 and collaboration with Dr. Carlos Velez and GE Global Research. This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request. The authors attest that all data for this study are included in the paper. Data provided by a third party are listed in Acknowledgment. No data, models, or codes were generated or used for this paper.

d =

jet diameter (m)

k =

thermal conductivity (W/(m K))

l =

length (m)

p =

pressure (atm)

s =

cell dimension (m)

t =

thickness (m)

u =

velocity (m/s)

x =

axial coordinate, downstream position (m)

y =

perpendicular coordinate, penetration depth (m)

z =

lateral coordinate (m)

$m˙$ =

mass flowrate (kg/s)

$q˙$ =

heat flux (W/m²)

$w˙$ =

reaction rate (g/(m³ s))

J =

momentum flux ratio

L =

turbulent length scale (m)

T =

temperature (K; °C)

X =

mole fraction

Y =

mass fraction

y⁺ =

wall criterion

Da =

Damköhler number

ppmvd =

parts per million on volumetric, dry basis

RST =

Reynolds Stress Turbulence

γ-Re_θ =

transition model

Δ =

difference

ɛ =

turbulent dissipation rate

ρ =

density

τ =

time scale

φ =

equivalence ratio

ω =

vorticity

axial =

axial stage

chem =

chemical

exp =

experimentally measured

global =

entire combustor

jet =

axial jet

main =

main burner stage

max =

maximum

turb =

turbulent