Search

Quantification of aleatoric uncertainties due to the inherent variabilities in operating conditions and fuel composition is essential for designing and improving premixers in dry low-emissions (DLE) combustion systems. Advanced stochastic simulation tools require a large number of evaluations in order to perform this type of uncertainty quantification (UQ) analysis. This task is computationally prohibitive using high-fidelity computational fluid dynamic (CFD) approaches such as large eddy simulation (LES). In this paper, we describe a novel and computationally efficient toolchain for stochastic modeling using minimal input from LES, to perform uncertainty and risk quantification of a DLE system. More specially, high-fidelity LES, chemical reactor network (CRN) model, beta mixture model, Bayesian inference and sequential Monte Carlo (SMC) are integrated into the toolchain. The methodology is applied to a practical premixer of low-emission combustion system with dimethyl ether (DME)/methane–air mixtures to simulate auto-ignition events at different engine conditions. First, the benchmark premixer is simulated using a set of LESs for a methane/air mixture at elevated pressure and temperature conditions. A partitioning approach is employed to generate a set of deterministic chemical reactor network (CRN) models from LES results. These CRN models are then solved at the volume-average conditions and validated by LES results. A mixture modeling approach using the expectation-method of moment (E-MM) is carried out to generate a set of beta mixture models and characterize uncertainties for LES-predicted temperature distributions. These beta mixture models and a normal distribution for DME volume fraction are used to simulate a set of stochastic CRN models. The Bayesian inference approach through SMC method is then implemented on the results of temperature distributions from stochastic CRN models to simulate the probability of auto-ignition in the benchmark premixer. The results present a very satisfactory performance for the stochastic toolchain to compute the auto-ignition propensity for a few events with a particular combination of inlet temperature and DME volume fraction. Characterization of these rare events is computationally prohibitive in the conventional deterministic methods such as high-fidelity LES.

Quantification of aleatoric uncertainties due to the inherent variabilities in operating conditions and fuel composition is essential for designing and improving premixers in dry low-emissions (DLE) combustion systems. Advanced stochastic simulation tools require a large number of evaluations in order to perform this type of uncertainty quantification (UQ) analysis. This task is computationally prohibitive using high-fidelity computational fluid dynamic (CFD) approaches such as large eddy simulation (LES). In this paper, we describe a novel and computationally-efficient toolchain for stochastic modelling using minimal input from LES, to perform uncertainty and risk quantification of a DLE system. More specially, high-fidelity LES, chemical reactor network (CRN) model, beta mixture model, Bayesian inference and sequential Monte Carlo (SMC) are integrated into the toolchain. The methodology is applied to a practical premixer of low-emission combustion system with dimethyl ether (DME)/methane-air mixtures to simulate autoignition events at different engine conditions. First, the benchmark premixer is simulated using a set of LESs for a methane/air mixture at elevated pressure and temperature conditions. A partitioning approach is employed to generate a set of deterministic chemical reactor network (CRN) models from LES results. These CRN models are then solved at the volume-average conditions and validated by LES results. A mixture modelling approach using the expectation-method of moment (EMM) is carried out to generate a set of beta mixture models and characterise uncertainties for LES-predicted temperature distributions. These beta mixture models and a normal distribution for DME volume fraction are used to simulate a set of stochastic CRN models. The Bayesian inference approach through Sequential Monte Carlo (SMC) method is then implemented on the results of temperature distributions from stochastic CRN models to simulate the probability of autoignition in the benchmark premixer. The results present a very satisfactory performance for the stochastic toolchain to compute the autoignition propensity for a few events with a particular combination of inlet temperature and DME volume fraction. Characterisation of these rare events is computationally prohibitive in the conventional deterministic methods such as high-fidelity LES.

Effective flame anchoring maximizes the low-emissions operating window of gas-turbine systems featuring closely spaced, multiple lean flames on single injector units. The complex flame-flow interactions inherent to such designs are not well understood, in particular, at part-load gas turbine conditions. In the present work, large eddy simulation (LES) was used to investigate flames in an optically accessible, confined, linear array of five industrial swirl injector elements. Well-anchored, partially attached and fully lifted flames were observed and three cases corresponding to these were selected for modeling. The dynamically thickened flame (DTF) model was used with a newly developed reduced chemical kinetics scheme that contains low temperature reaction pathways. Chemical explosive mode analysis (CEMA) shows that the flames exhibit thin ignition zones near their base, while a more distributed region of strongly positive modes occur downstream when a transition to partial blow-off takes place. Extinctions are frequent in this region and create broken reaction zones which seem to exhibit both autoignition as well as flamelet characteristics. Further decrease in reactivity leads to a fully lifted flame. It is clear that part-load conditions are best investigated by models that are combustion regime independent.

The minimization of auto-ignition risk is critical to the design of premixers of high power aeroderivative gas turbines as an increased use of highly reactive future fuels (for example, hydrogen or higher hydrocarbons) is anticipated. Safety factors based on ignition delays of homogeneous mixtures are generally used to guide the choice of a residence time for a given premixer. However, auto-ignition chemistry under aeroderivative conditions is fast (0.5–2 ms) and can be initiated within typical premixer residence times. The analysis of what takes place in this short period necessarily involves the study of low-temperature auto-ignition precursor chemistry, but precursors can change with fuel and local reactivity. Chemical explosive modes (CEMs) are a natural alternative to study this as they can provide a measure for auto-ignition risk by considering the whole thermochemical state in the framework of an eigenvalue problem. When transport effects are included by coupling the evolution of the chemical explosive modes to turbulence, it is possible to obtain a measure of spatial auto-ignition risk where both chemical (e.g., ignition delay) and aerodynamic (e.g., local residence time) influences are unified. In this article, we describe a method that couples large eddy simulation (LES) to newly developed, reduced auto-ignition chemical kinetics to study auto-ignition precursors in an example premixer representative of real life geometric complexity. A blend of pure methane and di-methyl ether (DME), a common fuel used for experimental auto-ignition studies, was transported using the reduced mechanism (38 species/238 reactions) under engine conditions at increasing levels of DME concentrations until exothermic auto-ignition kernels were formed. The resolution of species profiles was ensured by using a thickened flame model where dynamic thickening was carried out with a flame sensor modified to work with multistage heat release. This paper is outlined as follows: First, a reduced mechanism is constructed and validated for modeling methane as well as DME auto-ignition. Second, sensitivity analysis is used to show the need for chemical explosive modes. Third, the thickened flame model modifications are described and then applied to an example premixer at 25 bar/890 K preheat. The chemical explosive mode analysis closely follows the large thermochemical changes in the premixer as a function of DME concentrations and identifies where the premixer is sensitive and flame anchoring is likely to occur.

The minimization of autoignition risk is critical to the design of premixers of high power aeroderivative gas turbines as an increased use of highly reactive future fuels (for example, hydrogen or higher hydrocarbons) is anticipated. Safety factors based on ignition delays of homogeneous mixtures, are generally used to guide the choice of a residence time for a given premixer. However, autoignition chemistry at aeroderivative conditions is fast (0.5–2 milliseconds) and can be initiated within typical premixer residence times. The analysis of what takes place in this short period necessarily involves the study of low-temperature autoignition precursor chemistry, but precursors can change with fuel and local reactivity. Chemical Explosive Modes are a natural alternative to study this as they can provide a measure of autoignition risk by considering the whole thermochemical state in the framework of an eigenvalue problem. When transport effects are included by coupling the evolution of the Chemical Explosive Modes to turbulence, it is possible to obtain a measure of spatial autoignition risk where both chemical (e.g. ignition delay) and aerodynamic (e.g. local residence time) influences are unified. In this article, we describe a method that couples Large Eddy Simulation to newly developed, reduced autoignition chemical kinetics to study autoignition precursors in an example pre-mixer representative of real life geometric complexity. A blend of pure methane and dimethyl ether (DME), a common fuel used for experimental autoignition studies, was transported using the reduced mechanism (38 species / 238 reactions) at engine conditions at increasing levels of DME concentration until exothermic autoignition kernels were formed. The resolution of species profiles was ensured by using a thickened flame model where dynamic thickening was carried out with a flame sensor modified to work with multi-stage heat release. The paper is outlined as follows: First, a reduced mechanism is constructed and validated for modeling methane as well as di-methyl ether (DME) autoignition. Second, sensitivity analysis is used to show the need for Chemical Explosive Modes. Third, the thickened flame model modifications are described and then applied to an example premixer at 25 bar / 890K preheat. The Chemical Explosive Mode analysis closely follows the large thermochemical changes in the premixer as a function of DME concentration and identifies where the premixer is sensitive and flame anchoring is likely to occur.

The main motivation of this study is to investigate detailed NOx and CO formation in high-pressure dump combustor fired with lean premixed methane-air mixture using CFD-CRN hybrid approach. Further, this study is extended to investigate the effect of H2 enrichment on emission formation in the same combustor. Three-dimensional steady RANS CFD simulations have been performed using a Flamelet Generated Manifold (FGM) model in Simcenter STAR-CCM+ 2019.2 with the DRM22 reduced mechanism. The CFD simulations have been modelled along with all three heat transfers modes: conduction, convection and radiation. The conjugate heat transfer (CHT) approach and participating media radiation modelling have been used here. Initially, CFD simulations are performed for five lean equivalence ratios (ϕ = 0.43–0.55, Tinlet = 673 K, Vinlet = 40 m/s) of pure methane-air mixture operating at 5 bar. The exit temperature and flame-length are compared with available experimental data. The automatic chemical reactor network has been constructed from CFD data and solved using the recently developed reactor network module of Simcenter STAR-CCM+ 2019.2 in a single framework for each cases. It is found out that the CRNs up to 10,000 PSRs can provide adequate accuracy in exit NOx predictions compared to experiments for pure methane cases. The contribution of NOx formation pathway, changes from N2O intermediate to thermal NO as equivalence ratio increases. Further studies are performed for two equivalence ratios (ϕ = 0.43 and 0.50 to simulate the impact of H2 addition (up to 40% by volume) on NOx formation pathways and CO emission. It is found out here that the contribution from NNH pathway increases for leaner equivalence ratio cases (ϕ = 0.43), while thermal pathway slightly increases for ϕ = 0.50 with increase in H2 content from 0% to 40%.

Critical slowing down (CSD) is a phenomenon that is common to many complicated dynamical systems as they approach critical transitions/bifurcations. We demonstrate that pressure signals measured during the onset of thermoacoustic instabilities in a gas turbine engine test exhibit evidence of CSD well before the oscillation amplitude increases. CSD was detected through both the variance and the lag-1 auto-regressive coefficient in a rolling window of the pressure signal. Increasing trends in both metrics were quantified using Kendall's τ, and the robustness and statistical significance of the observed increases were confirmed. Changes in the CSD metrics could be detected several seconds prior to changes in the oscillation amplitude. Hence, real-time calculation of these metrics holds promise as early warning signals of impending thermoacoustic instabilities.

Finite-rate chemical effects at gas turbine conditions lead to incomplete combustion and well-known emissions issues. Although a thin flame front is preserved on an average, the instantaneous flame location can vary in thickness and location due to heat losses or imperfect mixing. Postflame phenomena (slow CO oxidation or thermal NO production) can be expected to be significantly influenced by turbulent eddy structures. Since typical gas turbine combustor calculations require insight into flame stabilization as well as pollutant formation, combustion models are required to be sensitive to the instantaneous and local flow conditions. Unfortunately, few models that adequately describe turbulence–chemistry interactions are tractable in the industrial context. A widely used model capable of employing finite-rate chemistry is the eddy dissipation concept (EDC) model of Magnussen. Its application in large eddy simulations (LES) is problematic mainly due to a strong sensitivity to the model constants, which were based on an isotropic cascade analysis in the Reynolds-averaged Navier–Stokes (RANS) context. The objectives of this paper are: (i) to formulate the EDC cascade idea in the context of LES; and (ii) to validate the model using experimental data consisting of velocity (particle image velocimetry (PIV) measurements) and major species (1D Raman measurements), at four axial locations in the near-burner region of a Siemens SGT-100 industrial gas turbine combustor.

Multiple, interacting flames in DLE systems can increase flame surface area and promote mixing of hot-products into the reactants — leading to an efficient usage of combustion volume and improved injector performance. An optically-accessible, confined, linear array of five swirl nozzles was recently built [1] to investigate flame dynamics and validate computational strategies. The present work focuses on modeling a dataset representative of lean gas turbine conditions, using a flamelet approach. A preheated (500K), premixed fuel-air mixture (ϕ = 0.55, Tflame = 1732K) at atmospheric pressure was injected through the swirlers at 40 m/s into a rectangular chamber. High-speed laser measurements of the flow (3 component velocity field from 10 kHz stereoscopic particle image velocimetry (S-PIV)) and flame (planar laser induced fluorescence of the hydroxyl radical (OH-PLIF)) were used for model validation. The objectives of this work: (1) Evaluate a flamelet-progress variable method based on flamelet-generated manifolds (FGM) and examine its sensitivity to models for micro (scalar dissipation) and large scale mixing (anisotropic RANS vs LES) and (2) Obtain insight into the velocity field and flame stabilization in an interacting system. Computations indicate that high-swirl nozzles produce bluff-body flames anchored to shear-layer vortices due to an arrested flow expansion. The anisotropic RANS turbulence model under-predicts the recirculation zone strength but predicts flow development and Reynolds stress profiles fairly well. While LES is more accurate overall, both models over-predict flow fluctuations in the transitional flow at the end of the recirculation bubble where flow becomes axially positive. The flamelet approach predicts the flame-shape and length correctly but over-predicts the reaction rate in-between swirlers. The effect of including a reactive SDR model is to significantly increase flame-flow interaction (higher scalar variance) but does not appear to influence the overall shape or location of the flame.

Sensitivity to stretch and differential diffusion of chemical species are known to influence premixed flame propagation, even in the turbulent environment where mass diffusion can be greatly enhanced. In this context, it is convenient to characterize flames by their Lewis number (Le), a ratio of thermal-to-mass diffusion. The work reported in this paper describes a study of flame stabilization characteristics when the Le is varied. The test data is comprised of Le ≪ 1 (Hydrogen), Le ≈ 1 (Methane), and Le > 1 (Propane) flames stabilized at various turbulence levels. The experiments were carried out in a Hot exhaust Opposed-flow Turbulent Flame Rig (HOTFR), which consists of two axially-opposed, symmetric turbulent round jets. The stagnation plane between the two jets allows the aerodynamic stabilization of a flame, and clearly identifies fuel influences on turbulent flames. Furthermore, high-speed Particle Image Velocimetry (PIV), using oil droplet seeding, allowed simultaneous recordings of velocity (mean and rms) and flame surface position. These experiments, along with data processing tools developed through this study, illustrated that in the mixtures with Le ≪ 1, turbulent flame speed increases considerably compared to the laminar flame speed due to differential diffusion effects, where higher burning rates compensate for the steepening average velocity gradient, and keeps these flames almost stationary as bulk flow velocity increases. These experiments are suitable for validating the ability of turbulent combustion models to predict lifted, aerodynamically-stabilized flames. In the final part of this paper, we model the three fuels at two turbulence intensities using the FGM model in a RANS context. Computations reveal that the qualitative flame stabilization trends reproduce the effects of turbulence intensity, however, more accurate predictions are required to capture the influences of fuel variations and differential diffusion.

Finite-rate chemical effects at gas turbine conditions lead to incomplete combustion and well-known emissions issues. Although a thin flame front is preserved on an average, the instantaneous flame location can vary in thickness and location due to heat losses or imperfect mixing. Post-flame phenomena (slow CO oxidation or thermal NO production) can be expected to be significantly influenced by turbulent eddy structures. Since typical gas turbine combustor calculations require insight into flame stabilization as well as pollutant formation, combustion models are required to be sensitive to the instantaneous and local flow conditions. Unfortunately, few models that adequately describe turbulence-chemistry interactions are tractable in the industrial context. A widely used model capable of employing finite-rate chemistry, is the Eddy Dissipation Concept (EDC) model of Magnussen. Its application in large eddy simulations (LES) is problematic mainly due to a strong sensitivity to the model constants which were based on an isotropic cascade analysis in the RANS context. The objectives of this paper are: (i) To formulate the EDC cascade idea in the context of LES; and (ii) To validate the model using experimental data consisting of velocity (PIV measurements) and major species (1-D Raman measurements), at four axial locations in the near-burner region of a Siemens SGT-100 industrial gas turbine combustor.

At a given power level, lean premixed (DLE) gas turbines vary equivalence ratio (ϕ) for optimal performance. This range is usually determined by variations in ambient conditions, acoustic response of the system, and emissions trade-off (e.g. between NOx and CO). In this work, the effects of ϕ variation on premixed jet flame lengths are investigated, by modeling the pressurized jet experiments of Griebel et al. [1]. While previous modeling of these experiments focused on a priori tabulated chemistry based methods, in this work we investigate an approach that represents finite-rate effects explicitly using skeletal chemistry (16 species, 41 reactions) in RANS and LES. Two equivalence ratios (ϕ = 0.56 and ϕ = 0.43) corresponding to the two extremes of flame lengths are chosen from the experimental database for 673K mixture preheat, 5 bar and 40 m/s jet velocity. A better correspondence with the experimentally measured flame length was achieved for ϕ = 0.43 than for ϕ = 0.56 indicating that the model is suitable when finite-rate effects are dominant but requires extensions for flames closer to the flamelet regime. It was found, further, that the RANS-EDC models failed to predict the confined turbulent jet development, as well as the flame lengths accurately, and demonstrated that scale resolution is required even for a relatively simple configuration.

CFD predictions of emissions such as NOx and CO in industrial lean-premixed gas turbine combustors depend heavily on the degree to which the complexity of turbulent mixing and turbulence-chemistry interaction in the flow-field is modeled. While there is much difficulty in obtaining detailed and accurate internal data from high pressure combustors, there is a definite need for accurately understanding the flow physics towards the improvement of design. This work summarizes some experience with using the RANS and LES approaches in a commercial code, Fluent 6.3, to predict CO emissions and temperature trends in the two-stage Rolls-Royce RB211-DLE combustor. The predictions are validated against exit emissions (obtained from exhaust gas analysis) and some thermal paint tests for qualitative agreement on flame-stabilization. The upstream geometry (plenum and counter-swirlers) was included in order to minimize the effect of boundary conditions on the combustion zone. The presumed pdf approach as well as finite-rate chemistry models using the eddy dissipation concept were used to compare the predictions. It was found that there was a very significant benefit in moving to more advanced turbulence modeling methods to obtain realistic predictions in a confined, swirling burner. Thermal paint tests indicated that flame stabilization and temperatures (and therefore CO) was incorrectly predicted in the RANS context. LES results, on the other hand, more accurately predicted flame stabilization with corresponding improvements in the exit CO predictions. Ongoing work focuses on the variations that can be expected by varying discretization schemes, combustion models and sub-grid turbulence models as well as obtaining detailed internal data suitable for LES comparisons.Copyright © 2010 by ASME