Sanjay M. Correa

A Review of NOx Formation Under Gas-Turbine Combustion Conditions

Abstract Gas turbines offer very high cycle efficiency, exceeding 50% in modern 250 MW-class combined-cycle units for power generation, as well as very low NOv in the lean premixed combustion mode with natural gas fuel. They also account for virtually all commercial aeropropulsion systems, in which case kerosene-based fuel is used. To meet future NOv and CO regulations, a higher level of understanding of turbulence, chemical kinetics and their interactions is required. NOv in particular has become a pacing consideration, although other constraints are present. Selected NOv data obtained at laboratory and machine conditions with gaseous fuel are reviewed here. Although the important chemical reactions cover a wide range in effective Damkohler number, the measure of turbulence-chemistry interactions, it appears that NOv is formed in a distributed zone manner. Equilibrium and superequilibrium effects can broaden the NOv-forming zones beyond the fine scales of turbulence, even in non-premixed flames. Pressure...

Superequilibrium and thermal nitric oxide formation in turbulent diffusion flames

Abstract Measurements and modeling of the formation of superequilibrium radicals and nitric oxide in atmospheric pressure turbulent jet diffusion flames are presented which quantify the influence of superequilibrium on thermal NO x formation. Variation of fuel gas compositions (CO/H 2 /N 2 , CO/H 2 /CO 2 , and CO/H 2 /Ar) permits partial separation of chemical and fluid mechanical effects. Superequilibrium OH radical concentrations are measured by single-pulse laser saturated fluorescence and NO and NO 2 concentrations by probe sampling and chemiluminescent detection. Four different types of probes were used to quantify probe sampling effects. In turbulent reaction zones, virtually all of the NO x in the flame occurred in the form of NO but far downstream of the flame nearly half of the NO x occurred as NO 2 . Thermal NO x maximized near stoichiometric flame zones; the rich shift observed by others may be a probe sampling artifact. In turbulent CO/H 2 /N 2 jet diffusion flames, both measurements and a nonequilibrium turbulent combustion model show that superequilibrium decreases average temperatures by 250K, increases average OH concentrations by a factor of 4–6, and increases thermal NO x formation principally by broadening the range of mixture fraction (both rich and lean) where thermal NO x is formed. Calculated increases in thermal NO x due to superequilibrium in turbulent CO/H 2 /N 2 jet diffusion flames are factors of 2.5 at 1 atm and 1.4 at 10 atm. The two-scalar pdf model predicts that thermal NO x yield is independent of Reynolds number in disagreement with previous experimental reports.

Computational models and methods for continuous gaseous turbulent combustion

Abstract Computational methods used to simulate turbulent reacting flow in combustors of complex shape are presented. The physical sub-models include either chemically equilibrated or partially-equilibrated (radical pool) models of varying complexity and laminar flamelet models. Limitations of turbulence models are discussed. Some results in non-premixed jet flames are reviewed showing the effects of non-equilibrium phenomena on major and minor species, temperature and pollutants. An algorithm applicable to recirculating flow in bodies with arbitrarily contoured boundaries is developed. Issues governing the choice of grid systems, discretization operators, and numerical solution procedures for recirculating flows are emphasized. Numerical results illustrate these issues. The method is demonstrated by application to a modern annular gas-turbine combustor.

Turbulence-chemistry interactions in the intermediate regime of premixed combustion☆

Abstract A numerical model for the partially stirred reactor (PaSR) is developed, and the effects of turbulence on NO, CO, and other quantities are computed. Turbulent mixing is accounted for by the “Interaction-by-Exchange-with-the-Mean” submodel. The PaSR is described by a system of (2Ns+1) × Np first-order coupled o.d.e.s in time, where Ns ≡ number of species, and Np ≡ number of particles. Combustion of a 50% CO 50% H 2 (by vol.) fuel premixed with air is considered, represented by 18 species and 43 reactions. In the limit of mixing frequency ω becoming small, the solutions tend to those of the plug flow as expected. NO and CO increase with mixing frequency. In the range of time scales relevant to turbulent combustion, say, 10 −4 s 1 ω −2 s, NO increases by a factor of about 2 as the mixing time becomes small enough to affect the concentration of oxyhydrogen radicals while CO increases by over an order of magnitude. These variations agree qualitatively with experimental data from turbulent combustors. In-combustor stirring clearly plays a large role even in premixed combustion. The algorithm converges to the perfectly stirred reactor solution at large mixing frequencies. The partial equilibrium model is found to be reasonable for CO H 2 fuels in the present range of conditions, and effects a computational speedup by a factor on the order of 100. Besides providing a useful combustion model, the PaSR provides a test-bed for mixing models, for simplified chemical schemes, and for algorithms intended for particle-tracking pdf transport models.

Measurements and modeling of a bluff body stabilized flame

Abstract An axisymmetric bluff body stabilized nonpremixed turbulent flame of 27.5% CO/32.3% H 2 /40.2% N 2 -in-air was investigated. The recirculation zone stabilized the flame and provided greater strain rates than possible in jet or even piloted-jet flames. Major species, density, and temperature were measured using a laser Raman scattering system, which was modified to operate in a chemiluminescent environment. The computational model was based on partial equilibrium in the radical pool, an assumed shape pdf over the two thermochemical variables required, and the k -ϵ turbulence model for closure of the density-weighted averaged Navier Stokes equations. The equations were solved in the elliptic form appropriate to recirculating flow. Enough grid was added to reduce the transverse cell Reynolds numbers to below two, ensuring second-order accurate and stable discretization of convection operators and so eliminating artificial diffusion. Mean properties such as density were obtained at each node by convolution with the joint pdf over the two thermochemical scalars. The k -ϵ turbulence model gave too rapid an initial decay. Agreement was encouraging on mixture fraction mean and variance, temperature, and species concentration fields. The bluff body provides an intensely turbulent flowfield for interactions with combustion chemistry, and is within the scope of numerical analysis. To improve the turbulence model, and to have a formalism that permits three or more scalars as required for hydrocarbon fuels, pdf transport methods should be merged with conventional solvers for the mean hydrodynamics.

Computation of Flow in a Gas Turbine Combustor

Abstract A methodology for computing steady turbulent combusting flow in combustors of complex shape is presented. Included is discussion of fully- or partially-equilibrated chemical kinetic models, the interaction of turbulence and combustion, grid systems, discretization operators and solution procedures for recirculating flows. Examples that demonstrate the influence of these issues are reviewed. A package of three-dimensional codes for grid generation and flow analysis-developed in the course of these studies-is applied to the flow in a sector of a modern annular gas-turbine combustor. Results are compared with available data. The study demonstrates the utility of modern computational methods and indicates directions for future work

Parallel simulations of partially stirred methane combustion

Premixed methane combustion in a partially stirred reactor (PaSR) is studied numerically. The effects of turbulent stirring rate on NO, CO, and other quantities are computed. The chemistry is represented by a “full” scheme (27 species, 77 reactions) in the baseline study. Turbulence is accounted for by the “IEM” (Interaction-by-Exchange-with-the-Mean) submodel. The PaSR is described by a system of (Ns + 1) × Np first-order coupled o.d.e.s in time, where Ns ≡ number of species, and Np ≡ number of particles. The model is well suited to parallel computers, without which the present study would not have been practical. The speedup over serial computers is essentially linear in the number of processors used, until the number of particles per processor becomes small enough (<10) to affect load balance. The conditions are 30 atm, 1200-K inlet temperature, 800-K equilibrium temperature rise, and 2-ms reactor residence time (in the PSR limit). In the PFR limit the flow just starts to ignite, while in the PSR limit temperatures are very near equilibrium. PaSR simulations are conducted in the range 100–5000 Hz (mixing frequency), and in each case converge to a stochastic steady state and span the PFR-PSR limits smoothly. The correlation of NO with particle age decreases as frequency increases, and is within expected limits. The OH levels are uniform to within a factor of two in this frequency range, which is consistent with the “distributed” OH structures observed in turbulent diffusion flames. Simulations with a 25-step “skeletal” scheme agreed well with the baseline study above 1,000 Hz, but are about 400 K low on mean temperature at 100 Hz. The corresponding four-step “reduced” scheme failed to ignite in all cases, suggesting a need for reduced schemes which do not assume that the radicals are in a chemically steady state.

Assessment of a partial-equilibrium/monte carlo model for turbulent syngas flames

Abstract Calculations and data for a turbulent jet flame of 40% CO, 30% H 2 , and 30% N 2 in coflowing air are compared extensively. The calculations are based on a partial-equilibrium model for the oxyhydrogen radical pool including CO, and on a velocity-composition joint probability density function (pdf), which closes the turbulent flux and mean chemical source terms. The pdf is joint between the three velocity components and two thermochemical scalars needed to describe partial-equilibrium conditions. The equation is solved numerically by a Monte Carlo technique. The data used are major species concentrations and temperature from pulsed Raman scattering. Difficulties with Raman measurements at high temperatures and of measuring CO 2 directly are discussed. The Raman signals are taken from previous studies but here are corrected for high-temperature effects and CO 2 vibrational spectra. Temperatures are obtained from the instantaneous density of the major species rather than from the Stokes/anti-Stokes ratio, which is more affected by chemiluminescence. The level of agreement between the model and the data is more favorable to the partial-equilibrium model than previously thought. The relative simplicity of the partial-equilibrium model makes it a candidate for practical calculations.

Prediction and measurement of a non-equilibrium turbulent diffusion flame

Superequilibrium radical concentrations in a turbulent CO/H2/N2 jet diffusion flame are computed using a two-scalar pdf model and directly measured using single pulse laser saturated OH fluorescence. The model is based on the averaged Navier-Stokes equations and the k∈l turbulence model. Non-equilibrium chemistry is accounted for by including CO in the partially equilibrated oxyhydrogen radical pool. Two scalars (mixture fraction and eaction progress suffice to describe the thermochemical system. Laser saturated fluorescence is used to directly measure the mean and fluctuating components of OH concentrations and thus the radical pool. Measurements and model both find mean OH concentrations which are four to six times larger than equilibrium with rms values of OH concentration also reasonably predicted. Superequilibrium effects are predicted to lower the mean temperature by as much as 250 K in agreement with experiments. Evidence of the breakdown of partial equilibrium was found in cool fuel-rich zones where predictions of temperature and OH concentration were too high. Extensions of the model to predict thermal NO formation and CO burnout are discussed.

A direct comparison of pair-exchange and IEM models in premixed combustion

Abstract Two pair-exchange mixing models, viz., the original Curl model and a modification thereof, are compared with the “Interaction-by-Exchange-with-the-Mean” (IEM) model, in the context of homogeneous combustion. The IEM model is attractive because it permits highly “parallelizable” computation, but the consequences of certain peculiarities—such as determinism and the shape-preserving relaxation of the initial pdf of a conserved scalar—need to be examined in the context of combustion. A numerical simulation of a partially stirred reactor (PaSR) is used to directly compare the three models, without the additional errors that contaminate comparisons made in simulations of flowfields. The fuel is 50%CO/50%H 2 (by vol.). The kinetic scheme consists of 11 species and 23 reactions. The stoichiometry of the premixed inflow leads to a PSR temperature of 1740 K, but to blowout in a PFR. The PaSR mixing frequency was varied in the range 10 Hz to 10 4 Hz, by factors of √10. The pair-exchange models predict blowout earlier than does the IEM model. Means computed from the IEM model are less noisy, a consequence of the determinism inherent in the IEM model. Despite these differences, the pdfs and scatterplots of temperature, CO, OH, and O—selected because of their intrinsic importance as well as their influence on NO x emissions—are very similar between the three models. The similarity increases with the mixing frequency, which is significant given that practical (e.g., gas-turbine) combustors operate at high mass-loadings and therefore necessarily at high mixing rates.