David T. Pratt

Characterization of NOx, N20, and CO for Lean-Premixed Combustion in a High-Pressure Jet-Stirred Reactor

A high-pressure jet-stirred reactor (HP-JSR) has been built and applied to the study of NO{sub x} and N{sub 2}O formation and CO oxidation in lean-premixed (LPM) combustion. The measurements obtained with the HP-JSR provide information on how NO{sub x} forms in lean-premixed, high-intensity combustion, and provide comparison to NO{sub x} data published recently for practical LPM combustors. The HP-JSR results indicate that the NO{sub x} yield is significantly influenced by the rate of relaxation of super-equilibrium concentrations of the O-atom. Also indicated by the HP-JSR results are characteristic NO{sub x} formation rates. Two computational models are used to simulate the HP-JSR and to provide comparison to the measurements. The first is a chemical reactor model (CRM) consisting of two perfectly stirred reactors (PSRs) placed in series. The second is a stirred reactor model with finite rate macromixing (i.e., recirculation) and micromixing. The micromixing is treated by either coalescence-dispersion (CD) or interaction by exchange with the mean (IEM) theory. Additionally, a model based on one-dimensional gas dynamics with chemical reaction is used to assess chemical conversions within the gas sample probe.

NOx Sensitivities for Gas Turbine Engines Operated on Lean-Premixed Combustion and Conventional Diffusion Flames

NOx exhaust emissions for gas turbine engines with lean-premixed combustors are examined as a function of combustor pressure (P), mean residence time (τ), fuel-air equivalence ratio (φ), and inlet mixture temperature (Ti). The fuel is methane. The study is accomplished through chemical reactor modeling of the combustor, using CH4 oxidation and NOx kinetic mechanisms currently available. The NOx is formed by the Zeldovich, prompt, and nitrous oxide mechanisms.The combustor is assumed to have a uniform φ, and is modeled using two reactors in series. The first reactor is a well-stirred reactor (WSR) operating at incipient extinction. This simulates the initiation and stabilization of the combustion process. The second reactor is a plug-flow reactor (PFR), which simulates the continuation of the combustion process, and permits it to approach completion. For comparison, two variations of this baseline model are also considered. In the first variation, the combustor is modeled by extending the WSR until it fills the whole combustor, thereby eliminating the PFR. In the second variation, the WSR is eliminated, and the combustor is treated as a PFR with recycle. These two variations do not change the NOx values significantly from the results obtained using the baseline model.The pressure sensitivity of the NOx is examined. This is found to be minimum, and essentially nil, when the conditions are P = 1 to 10atm, Ti = 600K, and φ = 0.6. However, when one or more of these parameters increases above the values listed, the NOx dependence on the pressure approaches P raised to a power of 0.4-to-0.6.The source of the NOx is also examined. For the WSR operating at incipient extinction, the NOx is contributed mainly by the prompt and nitrous oxide mechanisms, with the prompt contribution increasing as φ increases. However, for the combustor as a whole, the nitrous oxide mechanism predominates over the prompt mechanism, and for φ of 0.5-to-0.6, competes strongly with the Zeldovich mechanism. For φ greater than 0.6-to-0.7, the Zeldovich mechanism is the predominant source of the NOx for the combustor as a whole.Verification of the model is based on the comparison of its output to results published recently for a methane-fired, porous-plate burner operated with variable P, φ, and Ti. The model shows agreement to these laboratory results within a factor two, with almost exact agreement occurring for the leanest and coolest cases considered. Additionally, comparison of the model to jet-stirred reactor NOx data is shown. Good agreement between the model results and the data is obtained for most of the jet-stirred reactor operating range. However, the NOx predicted by the model exhibits a stronger sensitivity on the combustion temperature than indicated by the jet-stirred reactor data.Although the emphasis of the paper is on lean-premixed combustors, NOx modeling for conventional diffusion-flame combustors is presented in order to provide a complete discussion of NOx for gas turbine engines.Copyright

Coalescence/dispersion modeling of turbulent combustion in a jet-stirred reactor

The results of a direct simulation of high-intensit y methane/air combustion are compared to previously reported measurements of OH and O atom concentrations in a jet-stirred reactor. The simulation uses a zerodimensional coalescence/dispersion model that requires only the initial reactor feed conditions, an appropriate chemical rate mechanism, and estimates of the fluid mixing time scales. Empirical time scales, determined by forcing agreement between simulation and experiment, are of the same order of magnitude as those estimated from either global fluid dissipation length and velocity scales or inviscid arguments based on reactor length and velocity scales. Simulations using these estimates satisfactorily represent the measured temperature and species profiles over a wide range of fuel-air equivalence ratios and, in direct contrast to perfectly stirred reactor predictions, show close agreement with the measured values near stoichiometric conditions. Nomenclature c = concentration C/D = coalescence dispersion L = reactor length scale M = total mass flow rate N = number of Monte Carlo particles TV = time rate of change of TV TV5 = number of time series samples P(c) = concentration probability density function P0(c) = reactor inlet stream concentration probability density function pdf = probability density function PSR = perfectly stirred reactor

Effects of Incomplete Premixing on NOx Formation at Temperature and Pressure Conditions of LP Combustion Turbines

A probability density function/chemical reactor model (PDF/CRM) is applied to study how NOx emissions vary with mean combustion temperature, inlet air temperature, and pressure for different degrees of premixing quality under lean-premixed (LP) gas turbine combustor conditions. Inlet air temperatures of 550, 650 and 750 K, and combustor pressures of 10, 14 and 30 atm are examined in different chemical reactor configurations. Primary results from this study are: incomplete premixing can either increase or decrease NOx emissions, depending on the primary zone stoichiometry; an Arrhenius-type plot of NOx emissions may have promise for assessing the premixer quality of lean-premixed combustors; and decreasing premixing quality enhances the influence of inlet air temperature and pressure on NOx emissions.Copyright

Characterization of NOx, N2O, and CO for Lean-Premixed Combustion in a High-Pressure Jet-Stirred Reactor

A high-pressure jet-stirred reactor (HP-JSR) has been built and applied to the study of NOx and N2O formation and CO oxidation in lean-premixed (LPM) combustion. The measurements obtained with the HP-JSR provide information on how NOx forms in lean-premixed, high-intensity combustion, and provide comparison to NOx data published recently for practical LPM combustors. The HP-JSR results indicate that the NOx yield is significantly influenced by the rate of relaxation of super-equilibrium concentrations of the O-atom. Also indicated by the HP-JSR results are characteristic NOx formation rates. Two computational models are used to simulate the HP-JSR, and to provide comparison to the measurements. The first is a chemical reactor model (CRM) consisting of two perfectly-stirred reactors (PSRs) placed in series. The second is a stirred reactor model with finite rate macromixing (i.e., recirculation) and micromixing. The micromixing is treated by either coalescence-dispersion (CD) or interaction-by-exchange-with-the-mean (IEM) theory. Additionally, a model based on one-dimensional gas dynamics with chemical reaction is used to assess chemical conversions within the gas sample probe.Copyright

EXPONENTIAL-FITTED METHODS FOR STIFF ORDINARY DIFFERENTIAL EQUATIONS

It is well known that exponential functions are inherently superior to polynomials for representing the solution of damped or “stiff” systems of ordinary differential equations. However, the success of Gears method for automatically regulating the order of polynomial approximation, together with the introduction of backward-difference methods for the solution of stiff systems, has heretofore inhibited the development of the theoretical background required to enable the effective use of exponential functions as approximants for solving stiff systems of differential equations. A coherent strategy is presented for the computationally efficient estimation of local error and of local order of accuracy for exponential-based methods, which now makes it possible to achieve the same efficiency of automatic steplength control as that presently employed for linear multistep methods.

Supersonic flameholding by attached oblique shock waves

The object of this investigation was to explore various regimes of shock-induced combustion that may occur as a result of supersonic flow of a reactive gas mixture past a simple compression ramp or wedge. Attention was confined to combinations of wedge angles, fuel-air mixture ratios and freestream velocities that permit the resulting oblique shock or oblique detonation wave to remain attached. Computational fluid dynamic (CFD) simulations, both inviscid and viscous, were performed and analyzed in order to investigate two modes of compression ramp flameholding; namely, complete oblique detonation wave (ODW) and shock-induced combustion. In addition, the possibility of flame persistence or hysteresis was explored. It was preliminarily concluded that inclusion of viscous effects may be essential to accurately predict the occurrence of the various flameholding modes.

Conical detonation waves - A comparison of theoretical and numerical results

A predictive capability for conical oblique detonation waves (ODWs) is discussed with particular attention given to the special case of an axisymmetric conical projectile, which yields necessary conditions for attached waves. A multidimensional numerical model was used to study the influence of a finite Damkoehler number (Da) and viscosity on the formation of a conical ODW. The ODW is predicted for Da on the order of 10, and is found to be important in the ODW formation. For a Da of 1, shock induced combustion is predicted. Simulations of the experimental geometry showed general agreement with experimental flow features, but underpredicted the shock angle. It is suggested that this difference may be caused by inadequate boundary layer resolution, lack of grid orthogonality at the cone surface, and an inappropriate chemical mechanism.