Experimental and modeling investigation of the effect of air preheat on the formation of NOx in an RQL combustor
- First online:
The Rich-burn/Quick–mix/Lean-burn (RQL) combustor concept has been proposed to minimize the formation of oxides of nitrogen (NOx) in gas turbine systems. The success of this low-NOx combustor strategy is dependent upon the links between the formation of NOx, inlet air preheat temperature, and the mixing of the jet air and fuel-rich streams. Chemical equilibrium and kinetics modeling calculations and experiments were performed to further understand NOx emissions in an RQL combustor. The results indicate that as the temperature at the inlet to the mixing zone increases (due to preheating and/or operating conditions) the fuel-rich zone equivalence ratio must be increased to achieve minimum NOx formation in the primary zone of the combustor. The chemical kinetics model illustrates that there is sufficient residence time to produce NOx at concentrations that agree well with the NOx measurements. Air preheat was found to have very little effect on mixing, but preheating the air did increase NOx emissions significantly. By understanding the mechanisms governing NOx formation and the temperature dependence of key reactions in the RQL combustor, a strategy can be devised to further reduce NOx emissions using the RQL concept.
List of symbols
- A J /A M
Jet-to-mainstream area ratio
= n((d/2)/R)2 for a cylindrical duct
(S/H)(J)0.5 for a rectangular duct
= π(2*J)0.5/n for a cylindrical duct; derived in 
- C d
Orifice discharge coefficient
Round hole diameter
- d j
Effective round hole diameter = (d)(C d )0.5
Jet-to-mainstream density ratio, ρ J /ρ M
Duct height for a rectangular duct
= R for a cylindrical duct
Jet-to-mainstream momentum-flux ratio
= (ρV 2) jets /(ρU 2) main
= (MR)2 /DR/(C d )2 /(A J /A M )2
Jet-to-mainstream mass-flow ratio
= (ρVA total ) jets /(ρUA can cross-section ) main
= (DR)(V J /U M )(C d )(A J /A M )
Number of round holes in quick-mix module
Radius of the quick-mix module
Radial distance from the module center
- U M
Unmixed mainstream velocity
- V J
Jet exit velocity
Downstream distance; x = 0 at orifice leading edge
- ρ J
Density of jet flow
- ρ M
Density of mainstream flow
The Rich-burn/Quick-mix/Lean-burn (RQL) combustor concept was introduced in 1980 by Mosier et al.  as a strategy to reduce oxides of nitrogen (NOx) emissions from stationary gas turbine engines using fuels that were high in fuel-bound nitrogen. Later in the 1980′s, studies co-sponsored by the National Aeronautics and Space Administration (NASA) and the Environmental Protection Agency (EPA), continued [2–5] that were motivated by the RQL’s potential for having low NOx emissions for these fuels. A study of the effect on NOx emissions from fuels having high amounts of fuel-bound nitrogen was reported by Mao and Barth  in a staged combustor using CH3NH2 doping to simulate fuel-bound nitrogen.
In the 1990′s the RQL concept was selected by the High Speed Research (HSR) program at NASA as a candidate for the reduction of oxides of nitrogen in engines for a High Speed Civil Transport (HSCT) [7–11] in part because in aeroengine applications an RQL may be preferred over lean premixed prevaporized (LPP) options due to safety considerations and overall performance (e.g. stability) of the RQL throughout the duty cycle. Today, RQL technology is used in aeroengines deployed commercially.
In stationary applications, lean premixed combustor technology is the standard. Safety considerations are not as severe, the duty cycle is more constrained, and the reduction in NOx emissions is more substantial for LPP combustors in contrast to RQL technology. However, RQL combustor technology is of growing interest for stationary applications due to the attributes of more effectively processing fuels of complex and varying composition. The latter is becoming more important with the increasing international competition for fuels in general, and an increasing interest in liquefied natural gas, biomass fuels, and “opportunity fuels” (e.g., land-fill gases, digester gases, well-head gases) to either complement domestic fuel sources or serve as the sole source to a large region of a country (e.g. Southern California) or a country as a whole (e.g. South Korea).
1.1 RQL concept
The addition of oxygen in the mixing zone quickly converts the non-oxygen containing species in the Total Fixed Nitrogen (TFN) in the rich zone effluent to NOx , but is needed to oxidize the high concentrations of carbon monoxide, hydrogen, and hydrocarbon intermediates. If no additional NOx is formed in the mixing and lean zones, the NOx concentration at the combustor exit would be approximately equal to the sum of the mole fractions of the prompt-NO and the major non-oxygen containing species in the rich zone effluent times the rich-to-total mass-flow ratio. Thus, the goal in achieving low NOx in an RQL combustor is to mix air into the rich zone effluent quickly and in such a manner as to minimize the formation of thermal-NOx to achieve an exhaust at the exit of the combustor that is comprised of the major products of combustion (CO2, H2O, N2, O2) and a near-zero concentration of pollutants (e.g., NOx, CO, HC).
1.2 Jet mixing
To address the conditions particular to applications of jets in crossflow (JIC), numerous studies have been conducted to yield insight on jet structure and penetration, jet entrainment of crossflow fluid, and the flow field distributions resulting from jet mixing [12–20]. These studies have investigated single unbounded jets, both bounded and unbounded multiple jets, asymmetrically bounded jets, heated jets, a heated mainflow, reacting flows, or the doping of either the jets or mainflow with a tracer to allow the measurement of scalars in the flow downstream of the jet orifices. While a variety of jet orifice shapes have been studied (e.g., triangular, tear-drop, square, rectangular, bluff, slanted, and streamlined slots), no option has been identified that penetrates significantly farther or mixes faster than a round jet in an unbounded crossflow.
For a confined flow, as in a gas turbine combustor, design methods for both rectangular and cylindrical configurations have been identified at Cranfield and NASA. These are described by Lefebvre .
1.2.1 Non-reacting JIC mixing
In an RQL mixing section, the MR is much higher (~10×) than in the dilution zone of a conventional combustor. Since the density ratio DR (= ρ J /ρ M ) and momentum-flux ratio J (= (ρ J V J 2 )/(ρ Μ U M 2 )) are about the same for the two applications, the biggest difference between the jets in conventional and RQL combustors is orifice size.
While the cylindrical geometry has been extensively researched and is directly relevant to combustor can configurations, modern large annular combustor configurations have spawned investigations of jets in crossflow in rectangular geometries.
Insight on scalar distributions resulting from mixing of confined jets in crossflow stemmed from the studies conducted in the 1970’s and 1980’s to address the mixing behavior associated with dilution jets in conventional gas turbine combustors. The extensive non-reacting studies of rows of jets in crossflow in rectangular, annular, and cylindrical ducts that are summarized in Holdeman  were funded by NASA to evaluate geometrical features (e.g., orifice spacing, opposed rows of orifices, axial staging of orifices, orifice shape, flow area convergence, effects of flow-area curvature, and a non-uniform mainstream scalar) and operating conditions (e.g. density ratio, momentum-flux ratio, mass-flow ratio) with the goal to optimize the mixing in a gas turbine combustor. Later, the NASA JIC empirical model from  (without the curvature effects) was implemented in the Excel® spreadsheet reported in [15–18].
For one-side injection optimum jet penetration was observed when C was ~2.5; significant underpenetration was observed for C less than half of the optimum value; significant overpenetration was observed when C was more than double the optimum value. For opposed rows, corresponding C values are ~half these (i.e. Coptimum ~1.25; for significant underpenetration C is <0.63; for significant overpenetration C is over 2.5). For a rectangular or annular duct the jets can be injected from one side, or there can be opposed rows of aligned or staggered jets. For multiple jets in a cylindrical duct, injection is always one sided and the duct height [e.g. H in Eq. (2)] is the radius of the can.
Traditionally, studies such as Hatch et al.  have defined “optimal mixing” as the shortest distance downstream from the jet orifice needed to obtain a uniform profile for key scalars (e.g., temperature, species concentration). The hypothesis is that the optimal mixing defined in this manner will minimize the production of oxides of nitrogen. Due to the complex set of variables, a design of experiments statistical approach was used by Kroll et al.  to select optimum variables from among the multiple factors that can affect jet mixing. In addition to non-reacting experiments and use of design of experiments methods, empirical modeling and CFD have been effectively employed both independent of and in conjunction with experiments.
The studies NASA funded in the 1990’s focused on optimizing the mixing section in the Rich burn/Quick mix/Lean burn (RQL) combustor scheme in rectangular, annular, and cylindrical ducts for both non-reacting and reacting flows [19, 20]. These papers included design procedures for achieving the most rapid mixing for both cylindrical  and rectangular ducts .
1.2.2 Reacting JIC mixing
Reacting flow studies of jets in a crossflow have been conducted in both conventional and RQL combustor configurations to complement the non-reacting studies and assess the impact of heat release on the mixing processes [24–29].
For the RQL-motivated studies at atmospheric pressure by Leong et al. [28, 29], a mixture of propane and air was employed to generate a representative rich-burn effluent, and a specially designed “conditioning” section was used to create a uniform presentation (e.g., temperature, velocity, composition, and concentration) of the rich effluent to the Quick-mix section. The injection of the jet air into the rich effluent results in a combination of mixing and reaction between the rich-burn effluent and the jet air (Fig. 3). Measurements of temperature, species composition, and species concentration can then be made downstream of the jet orifices in order to establish the efficacy of mixing as a function of downstream distance.
The results from the reacting experiments reveal that distributions of a conserved scalar from non-reacting experiments generally provide a satisfactory description of the mixing of a conserved scalar for reacting jets in a crossflow. Overall, the jets in a cylindrical duct need to penetrate to approximately half of the can radius in order to maximize the mixing and avoid either under-penetration or over-penetration.
The studies by Leong et al. [28, 29] reported results for reacting flow experiments in a cylindrical duct at atmospheric pressure without inlet air preheat. A carbon balance technique similar to that of Jones et al.  was used by Leong et al. [28, 29] to obtain the equivalence ratio. Alternate methods for obtaining the local equivalence ratio in a reacting flow were investigated by Demayo et al. .
The first exploration of the relation between mixing and NOx emissions in an RQL-motivated study utilized the results of non-reacting experiments, and superimposed analytically the kinetics of NOx production . The results suggested that the best mixer might not minimize NOx emissions.
Experimental measurements under reacting conditions at elevated pressure were provided by Meisl et al. . The inlet air was preheated but the temperature was not varied, nor were the number of orifices in the mixing section varied. The study focused on the effect of varying pressure on NOx emissions. The authors of  noted the two well-known sources for the emissions of NOx, namely prompt- and thermal-NOx. The first of these is the formation of NO in the fuel-rich primary zone , a contribution that Meisl et al.  found was relatively insensitive to pressure. The second source is the formation of thermal NOx that may occur in the Quick-mix section which increases significantly with pressure. The study by Meisl et al.  also concluded that NOx emissions are minimized by optimal mixing.
A study by Vardakas et al.  extended the work reported by Leong et al. [28, 29] to address the effects on both mixing and emissions of varying inlet air preheat and varying the number of orifices. This study investigated configurations with 8, 12, 14, and 22 round holes at J = 57 to give over- to optimum to under-penetrating jets. The experiments were conducted at atmospheric pressure with the jet and main air preheated independently. Selected experimental results from Vardakas et al.  were published in Holdeman et al. .
The primary conclusion based on the experimental results was that mainstream air preheat increases NOx more that jet air preheat, and that an “optimum” mixer may not minimize NOx emissions.
The mixing section was modular and included a cylindrical section through which the mainstream effluent passed, and to which jet air was supplied from a surrounding plenum. The plenum for the jet air was fed by four equally-spaced air ports located toward the base of the plenum. A high-temperature steel flow-straightening device installed in the plenum conditioned and equally distributed the jet air entering the mixing module.
The mixing modules were 280 mm (11 in.) in length with inner and outer diameters of 80 mm (3.15 in.) and 85 mm (3.35 in.). The row of orifices was positioned with its centerline 115 mm (4.5 in.) downstream from the module entrance. An alumina-silica blend of ceramic fiber paper provided sealing between the module and the stainless steel mating surfaces to form the plenum for the air jets.
Recirculating heaters of 20 and 25 kW were utilized to supply the necessary preheat to the main and jet air lines respectively. The preheat temperatures were independently established and controlled.
Each planar grid consisted of 16 points spread over a sector that included two orifices (Fig. 5b). The points included one point located at the center of the duct, and five points along each of the arcs at r/R = 1/3, 2/3, and 1.
The measurement points along each arc were distributed such that two points were aligned with the center of the orifices and three were aligned with the midpoint between orifice centers for all cases.
Species concentration measurements were obtained by sampling through a water-cooled stainless-steel probe by routing the sample through a heated line connected to the emission analyzers. Water was condensed from the gas before the sample was analyzed by non-dispersed infrared (NDIR) analysis for CO and CO2, paramagnetic analysis for O2, flame ionization detection (FID) for total hydrocarbons, and chemiluminesence (CLD) for NOx. Data are reported as measured and were not “corrected” for either the ambient humidity or the expected water content of the flow.
The extractive emissions measurement protocol was conventional. The analyzers were standardized equipment which was checked frequently with span gases. As for the relation between what exists in the flow and what is measured; this is a classical question, and they are assumed to be the same when standard procedures are followed.
2.3 Probe design
2.4 Measurement uncertainty
A measure of the uncertainty of the data is given by analyzing the species measurements to determine the midplane and centerplane means and variances . In a perfect world, the 3 midplane measurements would be equal and the 2 centerplane measurements would be equal at a given radius, but they often aren’t. Their variability is a measure of both spatial and species uncertainty in the measurements. This is discussed further in .
2.5 Experimental conditions
The experiments were performed for a jet-to-mainstream momentum-flux ratio (J) of 57 and a mass-flow ratio (MR) of 2.5. The total effective area of the mixing module orifices is 9 cm2 (1.4 in.2). The ratio of the total effective jet area to the cross-sectional area of the duct is 0.18. The fuel-rich equivalence ratio and overall equivalence ratio are 1.66, and 0.45, respectively. The operating pressure for the system is one atmosphere.
Momentum–flux ratio, J
Mass-flow ratio, MR
Discharge coefficient, C d
(Total jet area)/(cross-sectional area of duct), AJ/AM
Non-preheated inlet air temperature
22 °C (72 °F)
Inlet preheat temperature
260 °C (500 °F)
These analyses supplement the experimental results reported in [35, 36] with chemical equilibrium and chemical kinetics calculations to gain insight into the formation of NOx in the rich , mixing, and lean (called the 2nd stage in ) stages.
The NASA Equilibrium Code  was used to perform chemical equilibrium calculations with the thermodynamic data base of Kee et al. . The modeling assumptions made were for fixed equivalence ratios and constant enthalpy.
Chemical kinetic calculations were performed using the SENKIN computer code, and the CHEMKIN III Subroutines [39, 40]. The chemical kinetic mechanism of Miller and Bowman  was used to describe the lower hydrocarbon combustion and NOx formation chemical kinetics along with the propane oxidation mechanism of Sundaram and Froment . The conditions of the model represented those of the experiment (e.g. temperature and stoichiometry in each zone) except that the model did not account for any effects due to mixing and dispersion.
4 Results and discussion
NOx data were obtained to investigate the effect of preheating jet and main air flows in the RQL combustor. Data for the 12 round-hole quick-mixing module are presented here as in [35, 36] for conditions without preheat and with both main and jet air preheated to the same temperature. Chemical equilibrium and kinetic modeling calculations were performed to gain insight into the NOx formation mechanisms governing the experiments and to compare calculated levels to measured NOx concentrations.
4.1 Chemical equilibrium
Mao and Barth  reported that the minimum in TFN shifted to larger rich zone ϕ’s and higher NO concentrations as the loading of CH3NH2 in the feed increased. As there is no fuel-bound nitrogen in the fuel used in our study, we would expect to have rich zone minimums in both the ϕ for the TFN minimum and the NO concentration there.
4.2 Chemical kinetics
The equivalence ratios used for the chemical kinetic modeling correspond to operating equivalence ratios for the fuel-rich (ϕ > 1) and fuel-lean (ϕ < !) zones of the RQL combustor. Chemical kinetic calculations were also performed for stoichiometric (ϕ = 1) conditions because of their importance to NOx formation in transitioning from fuel-rich to fuel-lean zones of the combustor which occurs in the quick-mix zone. The adiabatic flame temperature was used as a baseline case for each of the chemical kinetic simulations. In addition, two lower temperatures were simulated to account for various levels of heat loss that might occur in practical systems. Measurements of temperature that were made in the RQL combustor were consistently bounded by the chemical kinetic modeling conditions.
4.2.1 ϕ = 1.66 (Fuel-rich zone results)
4.2.2 ϕ = 1.00 (Stoichiometric results)
As air is introduced to the fuel-rich zone effluent in the quick-mix section of the RQL combustor there is a transition in overall equivalence ratio from rich (ϕ > 1) to lean (ϕ < 1). This transition does not occur instantaneously, but rather is characterized by finite mixing times resulting in spatial and temporal inhomogeneities and fluctuations between these two stoichiometries. During this mixing process stoichiometric or near-stoichiometric conditions and high temperatures can exist locally with a corresponding increase in potential for the formation of NOx and other fixed nitrogen species.
In Fig. 11, the rate of NO production under stoichiometric conditions is shown to be exceedingly fast. The other fixed nitrogen species which are significant for these conditions include nitrous oxide (N2O) and nitrogen dioxide (NO2), both with significantly lower concentrations compared to NO. NO forms very rapidly for all temperature conditions investigated. As with the rich conditions in Fig. 10, NO concentration levels for stoichiometric conditions shown in Fig. 11 are very dependent on temperature with mole fractions of over 1E − 03 (1,000 ppm) for the 2,364 K case, less than 1E − 03 (1,000 ppm) for the 2,215 K case, and approximately 1E − 04 (100 ppm) for the 2,103 K case, all obtained within 2 ms. The N2O and NO2 species are produced almost as quickly as the NO, but stay consistently below a mole fraction of 1E − 06 (1 ppm). The NO results show that the concentration level in stoichiometric regions can range from a mole fraction of 1E − 03 (1,000 ppm) to 1E − 04 (100 ppm) depending on temperature.
Although the thermal-NOx mechanism can produce substantial NOx in the quick-mix zone, the objective is to minimize it so the combustor NOx emissions are dominated by contributions from the prompt-NOx and fuel-NOx mechanisms.
4.2.3 ϕ = 0.45 (Fuel-lean zone results)
4.3 Experimental results
Figure 13(a) presents contour plots of the NOx distribution for conditions without preheat and with both streams preheated. These plots show the effect that jet penetration has on determining the locations of peak NOx concentration. The contour plots indicate that concentrations are high in the shear layers on both sides on the maximum jet penetration but the highest NOx concentrations occur near the wall of the module in the wakes of the jets.
The effect of main & jet air preheat on overall NOx emissions is presented in the bar chart in Fig. 13b and shows a large increase in overall NOx emissions with both streams preheated. The results of Vardakas et al.  showed a significantly larger increase in NOx emissions with both streams preheated than there was with no preheat and when only the jet air was preheated.
The variations of the averages for the cases shown vary from a low of ~10 ppm for the case without preheat to a high of ~25 ppm with both streams preheated.
NOx can be produced by several mechanisms (prompt-, fuel-, and thermal-NOx mechanisms), but when thermal NOx formation is minimized, the NOx is being produced primarily in the fuel-rich zone, and through conversion from other TFN species in the early mixing stage. The fuel-NOx mechanism can convert the fuel-rich zone total fixed nitrogen (TFN) species other than NO (e.g., NH3, HCN), which are the primary non-oxygen containing nitrogen species in the fuel-rich zone, to NOx with relatively high conversion . Although the chemical modeling does not take into account the effects of the mixing jets, the modeling results suggest that most of the NOx observed in the experimental data is formed either in the quick-mix zone or in the fuel-rich zone.
This information can be used to devise technique(s) to further reduce overall NOx formation in the RQL combustor. With preheat NOx formation in the fuel-rich zone is higher than optimal because the optimal stoichiometry for minimum TFN shifts to more fuel-rich conditions as temperatures increase. Temperature based control of the fuel-rich zone stoichiometry is suggested.
The increased dependence of overall NOx emissions on main air preheat suggests another optimization of the RQL concept. NOx could be reduced by increasing heat transfer from the fuel-rich zone to the fuel-lean zone. One strategy could be to pass the jet air through the fuel-rich zone combustor liner to transfer heat from this zone to the quick-mix and fuel-lean zones which do not have as significant a dependence upon preheat temperature.
Note that the completely mixed condition for a conserved scalar at MR = 2.5 would be ϕ = 0.45. Unmixed mainstream fluid would be ϕ = 1.66 and unmixed jet fluid would be ϕ = 0.
5 Summary and conclusions
An important feature in these results is the existence of a minimum for total fixed nitrogen (TFN) species.
Nitric oxide (NO) dominates the TFN species that are present for most equivalence ratios from ϕ = 0.4 to ϕ = 1.66.
Ammonia (NH3) and hydrogen cyanide (HCN) are increasingly important only in the fuel-rich regime.
Under air preheat conditions, the fuel-rich zone should be operated at a higher equivalence ratio to reduce TFN production in this zone.
NOx is formed in less than 2 ms at all conditions.
There exists sufficient residence time for significant amounts of NO to be formed in each zone.
An optimization of the RQL concept could include increases in the heat transfer from the fuel-rich zone to the mixing and fuel-lean zones and temperature-based control of fuel-rich zone stoichiometry.
These conclusions are based upon chemical equilibrium, chemical kinetics, and experimental results that point to primary production of NOx in the fuel-rich and early mixing zones. The mechanisms that lead to this NOx production and emissions are prompt-NOx and fuel-NOx. Thermal-NOx formed in the mixing zone can dominate emissions so NOx formation via this mechanism must be minimized in an RQL combustor.
The experimental work was supported by cooperation agreement NCC 3-412 from the NASA Glenn Research Center and the complementary chemical equilibrium and kinetics calculations were funded by the UCI Combustion Laboratory (UCICL) at the University of California, Irvine. The authors would also like to thank the Combustion Branch at the NASA Glenn Research Center for providing for the Open Access publication and color printing of this paper.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.