Modeling extinction and reignition in turbulent flames

The conditional moment closure method (CMC) has been extended to improve reactive species predictions in flames with significant local extinction and reignition. Simple first-order closure of the conditionally averaged reaction rate term does not give satisfactory results due to large fluctuations around the conditional mean and an alternative closure is suggested here. The new closure is based on a precomputed parameterized reference field that maps reactive species mass fractions as functions of mixture fraction and sensible enthalpy. During the computations, the reference field is continuously adjusted to ensure consistency with the CMC solution and doubly conditioned chemical source terms that are functions of time, space, mixture fraction, and sensible enthalpy can thus be obtained. Integration over sensible enthalpy space yields the improved singly conditioned chemical source term that can be used for the solution of the CMC equations. Full closure can be achieved by assuming a β-PDF for the probability distribution in sensible enthalpy space and an additional conditional variance equation needs to be solved. The overall agreement between the measured and the computed variance is satisfactory and the extended CMC model is applied to Sandia Flames D, E, and F. Excellent predictions of temperature, major species, intermediates, and NO are obtained in Flames D and E while temperature predictions can be significantly improved in Sandia Flame F.     

Assessment of closure schemes in second-order conditional moment closure against DNS with extinction and ignition

The performance of second-order conditional moment closure (CMC) depends on models to evaluate conditional variances and covariances of temperature and species mass fractions. In this paper the closure schemes based on the steady laminar flamelet model (SLFM) are validated against direct numerical simulation (DNS) involving extinction and ignition. Scaling is performed to reproduce proper absolute magnitudes, irrespective of the origin of mismatch between local flamelet structures and scalar dissipation rates. DNS based on the pseudospectral method is carried out to study hydrogen-air combustion with a detailed kinetic mechanism, in homogeneous, isotropic, and decaying turbulent media. Lewis numbers are set equal to unity to avoid complication of differential diffusion. The SLFM-based closures for correlations among fluctuations of reaction rate, scalar dissipation rate, and species mass fractions show good comparison with DNS. The variance parameter in lognormal PDF and the constants in the dissipation term have been estimated from DNS results. Comparison is made for the resulting conditional profiles from DNS, first-order CMC, and second-order CMC with correction to the most critical reaction step according to sensitivity analysis. Overall good agreement ensures validity of the SLFM-based closures for modeling conditional variances and covariances in second-order CMC.     

NO formation in counterflow partially premixed flames

An experimental and computational study of NO formation in low-strain-rate partially premixed methane counterflow flames is reported. For progressive fuel-side partial premixing the peak NO concentration increased and the NO distribution along the stagnation streamline broadened. New temperature-dependent emissivity data for a SiO₂-coated Pt thermocouple was used to estimate the radiation correction for the thermocouple, thus improving the accuracy of the reported flame temperature. Flame structure computations with GRIMech 3.00 showed good agreement between measured and computed concentration distributions of NO and OH radical. With progressive partial premixing the contribution of the thermal NO pathway to NO formation increases. The emission index of NO (EINO) first increased and then decreased, reaching its peak value for the level of partial premixing that corresponds to location of the nonpremixed reaction zone at the stagnation plane. The observation of a maximum in EINO at a level of partial premixing corresponding to the nonpremixed reaction zone at the stagnation plane seems to be a consistent feature of low (<20 s⁻¹)-strain-rate counterflow flames.     

湍流扩散火焰中对基于混合分数的湍流燃烧模型的数值研究

根据条件矩模型(CMC)和小火焰面模型在模型构建上的相似,针对具有不同大小雷诺数和湍流-化学相互作用特性的非预混湍流射流火焰,对这两种模型进行了数值研究和比较.湍流燃烧模型采用Lagrangian型非稳态小火焰模型(LFM)和径向加权积分的CMC模型,而在H2/N2火焰的数值研究中还考虑了稳态小火焰模型的数值模拟结果....     

Premixed and nonpremixed generated manifolds in large-eddy simulation of Sandia flame D and F

Premixed and nonpremixed flamelet-generated manifolds have been constructed and applied to large-eddy simulation of the piloted partially premixed turbulent flames Sandia Flame D and F. In both manifolds the chemistry is parameterized as a function of the mixture fraction and a progress variable. Compared to standard nonpremixed flamelets, premixed flamelets cover a much larger part of the reaction domain. Comparison of the results for the two manifolds with experimental data of flame D show that both manifolds yield predictions of comparable accuracy for the mean temperature, mixture fraction, and a number of chemical species, such as CO₂. However, the nonpremixed manifold outperforms the premixed manifold for other chemical species, the most notable being CO and H₂. If the mixture is rich, CO and H₂ in a premixed flamelet are larger than in a nonpremixed flamelet, for a given value of the progress variable. Simulations have been performed for two different grids to address the effect of the large-eddy filter width. The inclusion of modeled subgrid variances of mixture fraction and progress variable as additional entries to the manifold have only small effects on the simulation of either flame. An exception is the prediction of NO, which (through an extra transport equation) was found to be much closer to experimental results when modeled subgrid variances were included. The results obtained for flame D are satisfactory, but despite the unsteadiness of the LES, the extinction measured in flame F is not properly captured. The latter finding suggests that the extinction in flame F mainly occurs on scales smaller than those resolved by the simulation. With the presumed β-pdf approach, significant extinction does not occur, unless the scalar subgrid variances are overestimated. A thickened flame model, which maps unresolved small-scale dynamics upon resolved scales, is able to predict the experimentally observed extinction to some extent.     

Experiments on the scalar structure of turbulent CO/H2/N2 jet flames

Scalar and velocity measurements are reported for two turbulent jet flames of CO/H₂/N₂ (40/30/30 volume percent) having the same jet Reynolds number of 16,700 but different nozzle diameters (4.58 mm and 7.72 mm). Simultaneous measurements of temperature, the major species, OH, and NO are obtained using the combination of Rayleigh scattering, Raman scattering, and laser-induced fluorescence. Three-component laser-Doppler velocimetry measurements on the same flames were performed at ETH Zurich and are reported separately. This paper focuses on the scalar results but includes some limited velocity data. Axial profiles of mixture fraction, major species mole fractions, and velocity in these two flames are in close agreement when streamwise distance is scaled by nozzle diameter. However, OH mole fractions are lower and NO mole fractions are higher near the stoichiometric flame length in the larger flame due to the lower scalar dissipation rates and longer residence times. Turbulent flame measurements are compared with steady strained laminar flame calculations. Laminar calculations show remarkably close agreement with measured conditional means of the major species when all diffusivities are set equal to the thermal diffusivity. In contrast, laminar flame calculations that include the normal Chemkin treatment of molecular transport are clearly inconsistent with the measurements. These results suggest that turbulent stirring has a greater influence than molecular diffusion in determining major species concentrations at the flow conditions and locations considered in the present experiments, which begin at an axial distance of 20 nozzle diameters. Analysis of the conditional statistics of the differential diffusion parameter supports this conclusion, though some evidence of differential diffusion is observed. With regard to validation of turbulent combustion models, this data set provides a target that retains the geometric simplicity of the unpiloted jet flame in coflow, while including a chemical kinetic system of intermediate complexity between hydrogen flames and the simplest hydrocarbon flames. Aspects of the measurements, including Favre-averaged profiles, conditional statistics, mixture fraction pdf’s, and departures from partial equilibrium, are presented and discussed in terms or their relevance to the testing of turbulent combustion submodels. The complete data are available on the World Wide Web for use in model validation studies.     

Using the tabulated diffusion flamelet model ADF-PCM to simulate a lifted methane-air jet flame

Two formulations of a turbulent combustion model based on the approximated diffusion flame presumed conditional moment (ADF-PCM) approach [J.-B. Michel, O. Colin, D. Veynante, Combust. Flame 152 (2008) 80-99] are presented. The aim is to describe autoignition and combustion in nonpremixed and partially premixed turbulent flames, while accounting for complex chemistry effects at a low computational cost. The starting point is the computation of approximate diffusion flames by solving the flamelet equation for the progress variable only, reading all chemical terms such as reaction rates or mass fractions from an FPI-type look-up table built from autoigniting PSR calculations using complex chemistry. These flamelets are then used to generate a turbulent look-up table where mean values are estimated by integration over presumed probability density functions. Two different versions of ADF-PCM are presented, differing by the probability density functions used to describe the evolution of the stoichiometric scalar dissipation rate: a Dirac function centered on the mean value for the basic ADF-PCM formulation, and a lognormal function for the improved formulation referenced ADF-PCMχ. The turbulent look-up table is read in the CFD code in the same manner as for PCM models. The developed models have been implemented into the compressible RANS CFD code IFP-C3D and applied to the simulation of the Cabra et al. experiment of a lifted methane jet flame [R. Cabra, J. Chen, R. Dibble, A. Karpetis, R. Barlow, Combust. Flame 143 (2005) 491-506]. The ADF-PCMχ model accurately reproduces the experimental lift-off height, while it is underpredicted by the basic ADF-PCM model. The ADF-PCMχ model shows a very satisfactory reproduction of the experimental mean and fluctuating values of major species mass fractions and temperature, while ADF-PCM yields noticeable deviations. Finally, a comparison of the experimental conditional probability densities of the progress variable for a given mixture fraction with model predictions is performed, showing that ADF-PCMχ reproduces the experimentally observed bimodal shape and its dependency on the mixture fraction, whereas ADF-PCM cannot retrieve this shape.     

Direct numerical simulations of the double scalar mixing layer: Part II: Reactive scalars

The reacting double scalar mixing layer (RDSML) is investigated as a canonical multistream flow and a model problem for simple piloted diffusion flames. In piloted diffusion flames, the reacting fuel and oxidizer streams are initially separated by a central pilot stream at stoichiometric composition. The primary purpose of this pilot is to delay the mixing of the pure streams until a stable flame base can develop. In such multistream systems, the modeling of turbulent scalar mixing is complicated by the multiple feed streams, leading to more complex fine-scale statistics, which remain as yet an unmet modeling challenge compared to the simpler two-feed system. In Part I we described how multimodal mixture fraction probability density functions (PDFs) and conditional scalar dissipation rates can be modeled with a presumed mapping function approach. In this work we present an efficient and robust extension of the modeling to a general multistream reacting flow and compare predictions to three-dimensional direct numerical simulations (DNS) of the RDSML with a single-step reversible chemistry model and varying levels of extinction. With high extinction levels, the interaction with the pilot stream is described. Additionally, state-of-the-art combustion modeling calculations including conditional moment closure (CMC) and stationary laminar flamelet modeling (SLFM) are performed with the newly developed mixing model. Excellent agreement is found between the DNS and modeling predictions, even where the PDF is essentially a triple-delta shape near the flame base, so long as extinction levels are moderate to low. The suggested approach outlined in this paper is strictly valid only for flows that can be described by a single mixture fraction. For these flows the approach should provide engineers with fine-scale models that are of accuracy comparable to those already available for binary mixing, at only marginally higher complexity and cost.     

Imaging and radiation of laminar jet diffusion flames with low coflow air velocity in microgravity

We present imaging results and radiation measurements from laminar jet diffusion flames burning in coflowing air conditions. Color and pseudocolor flames are obtained and used to analyze flame brightness and shape, which show that flames under normal gravity are brighter than in microgravity. The longer residence times for microgravity flames result in increased radiative loss, which leads to local extinction and low temperature at the flame tip. Flame radiation fractions for microgravity flames are larger than those in normal gravity for C₂H ₄ and CH ₄. The velocity of coflowing air has a much more pronounced effect on radiation from microgravity flames compared to those in normal gravity. The radiation fractions from ethylene‐fueled flames in microgravity are large, leading to local extinction at the flame tip. We also analyzed the flame radiation fraction.     

Implementation of a dynamic thickened flame model for large eddy simulations of turbulent premixed combustion

A dynamic version of the thickened flame model for large eddy simulations (TFLES) of turbulent premixed combustion is implemented by coupling two independent codes through MPI (Message Passing Interface) running on massively parallel machines. The combustion code, AVBP, solves the usual filtered Navier–Stokes, mass fractions and energy balance equations on unstructured meshes while a second code takes advantage of the knowledge of resolved flow structures to automatically adjust the combustion model parameter from filtered resolved flow fields. This approach is tested against data from the F1, F2 and F3 pilot stabilised jet-flames experimentally investigated by Chen and coworkers. The model parameters determined dynamically are quite different for these cases, evidencing the decisive advantage of the dynamic procedure while statistical results are in good agreement with experimental data. The extra cost induced by the dynamic model is here about 20% but further optimisation remains possible.     

Evaluation of chemical effects of added CO2 according flame location

A numerical study has been conducted to clearly grasp the impact of chemical effects caused by added CO₂ and of flame location on flame structure and NO emission behaviour. Flame location affects the major source reaction of CO formation, CO₂+H→CO+OH and the H‐removal reaction, CH₄+H→CH₃+H₂. It is, as a result, seen that the reduction of maximum flame temperature due to chemical effects for fuel‐side dilution is mainly caused by the competition of the principal chain branching reaction with the reaction, CH₄+H→CH₃+H₂, while that for fuel‐side dilution is attributed to the competition of the principal chain branching reaction with the reaction, CO₂+H→CO+OH. The importance of the NNH mechanism for NO production, where the reaction pathway is NNH→NH→HNO, is recognized. In C‐related reactions most of NO is the direct outcome of (R171) and the contribution of (R171) becomes more and more important with increasing amount of added CO₂ as much as the reaction step (R171) competes with the key reaction of thermal mechanism, (R237), for N atom. This indicates a possibility that NO emission in hydrogen flames diluted with CO₂ shows less dependent behaviour upon flame temperature. Copyright © 2004 John Wiley & Sons, Ltd.     

CO2稀释对甲烷-高温空气扩散燃烧及NO生成特性影响的化学动力学分析

基于GRI-Mech 3.0详细化学反应机理,利用OPPDIF Code研究了CO2稀释比、预热温度及拉伸率对甲烷-高温空气层流对冲扩散火焰温度、热释放率、组分摩尔分数及NO生成特性的影响.研究结果表明,CO2稀释助燃空气能有效降低火焰中H、O及OH等基团摩尔分数,抑制燃烧过程链传播及链引发反应,从而减缓CH4氧化速率.随着助燃空气中CO2稀释比的增加,火焰最高温度逐渐降低,主氧化区及第二氧化区放热峰值变小,燃烧反应高温区变窄,NO生成指数E显著降低.当稀释比大于20%时,热力型NO随助燃空气温度升高规律并不明显.随着CO2稀释比的增加,快速型NO对NO生成量影响逐渐增强,成为高CO2稀释比下甲烷-高温空气扩散燃烧NO生成的主要路径.     

Investigation of the effect of pilot burner on lean blow out performance of a staged injector

The staged injector has exhibited great potential to achieve low emissions and is becoming the preferable choice of many civil airplanes. Moreover, it is promising to employ this injector design in military engine, which requires most of the combustion air enters the combustor through injector to reduce smoke emission. However, lean staged injector is prone to combustion instability and extinction in low load operation, so techniques for broadening its stable operation ranges are crucial for its application in real engine.
In this work, the LBO performance of a staged injector is assessed and analyzed on a single sector test section.The experiment was done in atmospheric environment with optical access. Kerosene-PLIF technique was used to visualize the spray distribution and common camera was used to record the flame patterns. Emphasis is put on the influence of pilot burner on LBO performance. The fuel to air ratios at LBO of six injectors with different pilot swirler vane angle were evaluated and the obtained LBO data was converted into data at idle condition. Results show that the increase of pilot swirler vane angle could promote the air assisted atomization, which in turn improves the LBO performance slightly. Flame patterns typical in the process of LBO are analyzed and attempts are made to find out the main factors which govern the extinction process with the assistance of spray distribution and numerical flow field results. It can be learned that the flame patterns are mainly influenced by structure of the flow field just behind the pilot burner when the fuel mass flow rate is high; with the reduction of fuel, atomization quality become more and more important and is the main contributing factor of LBO. In the end of the paper,conclusions are drawn and suggestions are made for the optimization of the present staged injector.     

初始压力对天然气-氢气-空气混合气火焰传播特性的影响

使用定容燃烧弹研究了不同初始压力下天然气-氢气-空气混合气的火焰传播规律,得到了初始压力、掺氢比和燃空当量比对无拉伸层流燃烧速率、质量燃烧流量的影响,结合高速纹影图片分析了影响火焰稳定性的因素(马克斯坦长度、火焰面两侧密度比和火焰厚度).结果表明,掺氢天然气无拉伸层流燃烧速率以及火焰的不稳定性受掺氢比、初始压力和燃空当量比的综合影响.结合高速纹影图片,得出火焰的稳定性会随初始压力的增加而减小;在相同的燃空当量比和掺氢比下,初始压力对密度比的影响不大,但是对火焰厚度的影响比较明显.     

Simulation of combustion in spark-ignition engine fuelled with natural gas-hydrogen blends combined with EGR

A numerical simulation of the influence of different hydrogen fractions, excess air ratios and EGR mass fractions in a spark-ignition engine was conducted. Good agreement between the calculated and measured incylinder pressure traces as well as pollutant formation trends was obtained. The simulation results show that NO concentration has an exponential relationship with temperature and increases sharply as hydrogen is added. EGR introduction strongly influences the gas temperature and NO concentration in the cylinder. The difference in temperature will lead to even greater difference in NO concentration. Thus, EGR can effectively decrease NO concentration. NO concentration reaches its peak value at the excess air ratio of 1.1 regardless of EGR mass fraction. The study shows that natural gas-hydrogen blend combined with EGR can realize a stable combustion and low NO emission in a spark-ignition engine.     

Numerical investigation for combustion characteristics of vacuum residue (VR) in a test furnace

It has become inevitable to search for alternative fuels due to current worldwide energy crisis. In this paper combustion characteristics of vacuum residue (VR) is investigated numerically against experimental data in typical operating conditions of a furnace. Heat release reaction is modeled as sequential steps of devolatilization, simplified gas phase reaction and char oxidation as for pulverized coal. Thermal and fuel NO are predicted by the conditional moment closure (CMC) method for estimation of elementary reaction rates. It turns out that Sauter mean diameter (SMD) of VR droplets is a crucial parameter for better combustion efficiency and lower NO. Reasonable agreement is achieved for spatial distributions of major species, temperature and NO for all test cases with different fuel and steam flow rates.     

Pressure dependence of NO formation in laminar fuel-rich premixed CH4/air flames

Effects of pressure on NO formation in CH₄/air flames at a fixed equivalence ratio of 1.3 are investigated. The axial profiles of temperature, OH, CH, and NO mole fractions are measured using laser-induced fluorescence and compared with one-dimensional flame calculations. The measured and calculated temperature, CH, and NO profiles in free flames are observed to vary upon increasing the pressure from 40 to 75 Torr, following a scaling law derived for a chemical mechanism containing only second-order reactions. At pressures 300–760 Torr, the measurements and calculations in burner-stabilized flames show increasing flame temperature and NO mole fractions when the mass flux is increased linearly with pressure, while the CH profiles remain unchanged. The observed deviation from the scaling law in the temperature profiles arises from the increasing contribution of three-body reactions to the flame front propagation velocity, leading to a decrease in the degree of burner stabilization. The deviation from the pressure scaling law for the NO mole fractions is due to the temperature dependence of the rate coefficient for the reaction between CH and N₂ and the fact that the temperature profiles themselves do not scale. In contrast, the surprisingly good scaling of the CH mole fractions with pressure indicates the dominant role of two-body reactions participating in the chain of chemical reactions leading to CH formation. The calculations using GRI-Mech 3.0 substantially overpredict (up to 50%) the measured nitric oxide concentrations for all pressures studied. The observed differences in the NO mole fraction may be addressed by improving the CH prediction.