Simulation Of Turbulent Combustion Research Articles

Including detailed chemistry features in the modeling of emerging low-temperature reactive flows: A review on the application to diluted and MILD combustion systems

Developing and optimizing new reactive systems with carbon-neutral fuels like biofuels, e-fuel, hydrogen, or ammonia is crucial for sustainable energy. This requires advanced technologies capable of fuel flexibility, high efficiency, and minimal pollutant emissions. However, these energy carriers still produce pollutants, especially NOx. To address this, engineers aim to lower combustion process temperatures by adopting different strategies such as burned gas recirculation, staging or increasing the air-to-fuel ratio. Yet, lean flames, though effective at emission reduction, are prone to instability and extinction, posing safety and mechanical risks. Emerging technologies like MILD Combustion, based on burned gas recirculation and reactant dilutions offer interesting solutions. The review article begins by synthesizing experimental studies and numerical simulations of MILD turbulent combustion. It then explores fundamental phenomena specific to diluted combustion (where MILD regimes are included as sub-sets), including autoignition and flame propagation. Using high-fidelity simulations and advanced experiments, it examines flow and mixing roles in reactive zones stabilization. Moving forward, the review paper addresses the inclusion of detailed chemical properties in modeling turbulent combustion systems. Scientific challenges revolve around modeling the intricate interactions between combustion chemistry and flow turbulence while maintaining computational efficiency compatible with industrial constraints. To address this, various simplified chemistry methods – such as reduced, tabulated, or optimized chemistry – have been developed. Additionally, turbulence/chemistry coupling modeling remains unresolved in simulations, with three main routes – geometrical, statistical, or reactor-based approaches – available for turbulent combustion modeling. The state-of-the-art in simplified chemistry and turbulent combustion modeling for low-temperature regimes is then focused on capturing MILD regimes, where there is a crucial impact of dilution by burnt gases, heat transfer, and turbulence mixing on the chemical flame structure. Recent advancements enabled by machine learning and deep learning algorithms are also highlighted. Lastly, the article underscores the critical need for data to validate models, emphasizing the importance of scale-bridging experiments.

A deep learning framework for supersonic turbulent combustion

The rapid simulation of supersonic turbulent combustion is a significant demand in scientific research and engineering applications for hypersonic vehicles. This paper proposes a deep learning framework for fast predicting unsteady turbulent combustion flow fields within the combustor of hypersonic vehicles. Based on convolutional neural networks and recurrent neural networks, this framework extracts spatial distribution characteristics of the flow fields and temporal evolution rules. And we enhance the traditional mean square error loss function by assigning loss weights to different channel data. Numerical simulations are conducted on the model scramjet combustor with various geometric structures to generate the dataset for training, and part of the untrained cases are used to verify the effectiveness. The results show that the proposed model, under different geometric structures, achieves high computational accuracy, with a correlation coefficient between the predicted results and the true values above 0.99. Considering the time cost of data transferring between heterogeneous systems, the model takes only 30 s to complete the calculation, representing an acceleration of at least two orders of magnitude compared to computational fluid dynamics. In the future, it can be applied to the rapid prediction of hypersonic vehicle performance and efficiently guide the optimal design of aircraft.

Implications of Using Scalar Forcing to Sustain Reactant Mixture Stratification in Direct Numerical Simulations of Turbulent Combustion

A recently proposed scalar forcing scheme that maintains the mixture fraction mean, root-mean-square and probability density function in the unburned gas can lead to a statistically quasi-stationary state in direct numerical simulations of turbulent stratified combustion when combined with velocity forcing. Scalar forcing alongside turbulence forcing leads to greater values of turbulent burning velocity and flame surface area in comparison to unforced simulations for globally fuel-lean mixtures. The sustained unburned gas mixture inhomogeneity changes the percentage shares of back- and front-supported flame elements in comparison to unforced simulations, and this effect is particularly apparent for high turbulence intensities. Scalar forcing does not significantly affect the heat release rates due to different modes of combustion and the micro-mixing rate within the flame characterised by scalar dissipation rate of the reaction progress variable. Thus, scalar forcing has a significant potential for enabling detailed parametric studies as well as providing well-converged time-averaged statistics for stratified-mixture combustion using Direct Numerical Simulations in canonical configurations.

Adaptive mesh refinement towards optimized mesh generation for large eddy simulation of turbulent combustion in a typical micro gas turbine combustor

Over the last years, several advanced micro Gas Turbine (mGT) cycle developments have been proposed, aiming at making the mGT more fuel and operational flexible. However, accurate data on real industrial combustors, assessing the performances and emissions of the combustion under unconventional diluted conditions or fuels involved in these novel cycles, are still missing. In this framework, Large Eddy Simulations, who allow to accurately assess the unsteady effects coupled to turbulent-chemistry interaction of reacting flows, offer an opportunity to better assess the combustion behaviour under these specific conditions. However, the computational cost remains much higher compared to RANS simulations. The mesh generation process might be complex, especially when the region of interest is not intuitively known. Therefore, Adaptive Mesh Refinement (AMR) methods allow the mesh to be refined only in the region where finer cells are required to capture important phenomena, i.e. in the flame front. By dynamically refining the mesh all along the simulation, the mesh is optimized in terms of cell quantity and distribution for more accurate results at potentially lower computational costs. In this work, an adaptive mesh refinement method based on a predefined criterion is successfully applied to the LES of a typical industrial mGT combustor, the Turbec T100. The results show that the adaptation strategy allows to automatically generate a dynamic mesh that is able to capture correctly the flame over time for an increased computational cost of 15%. Therefore, the human meshing effort is reduced while the automatic meshing leads to an optimized mesh.

Adaptive detached eddy simulation of turbulent combustion with the subgrid dissipation concept

Detached eddy simulation has become a widely used method in eddy simulations due to its balance between cost and accuracy. The recently developed subgrid dissipation concept (SDC) combustion model [Liu et al., “On the subgrid dissipation concept for large eddy simulation of turbulent combustion,” Combust. Flame 258, 113099 (2023)] is found to be more reasonable and accurate than the conventional eddy dissipation concept model in large eddy simulation (LES). In this paper, the SDC model is adapted to the ℓ2-ω adaptive detached eddy simulation framework, named DES-SDC. The required key quantities, including the fine structure mass fraction and dissipation rate, are appropriately blended across Reynolds-averaged Navier–Stokes and LES regions. The DES-SDC approach is validated using premixed bluff body stabilized flame, non-premixed swirl flame, and premixed swirl flame with complex geometry. It is much more tolerant to coarse mesh resolution than pure LES, yet it preserves the capability of resolving the key unsteady feature critical for the combustion process, as it is designed to be. The DES-SDC approach is relatively insensitive to the grid resolution. The present research provides a promising approach for accurately simulating practical unsteady turbulent combustion problems at an affordable computational cost.

Mathematical Simulation of Turbulent Combustion of Air–Propane Mixture in Swirl Flow

Mathematical Simulation of Turbulent Combustion of Air–Propane Mixture in Swirl Flow

Brief Review of Recent Achievements in the Flamelet Manifold Selection and Probability Density Distribution for Flamelet Manifold Variables

Abstract The flamelet model is a commonly used tool for turbulent combustion simulations in the engineering field due to its computational efficiency and compatibility with complex chemical reaction mechanisms. Despite being widely used for decades, the flamelet model still faces challenges when applied to complex flame configurations, such as partially premixed flames, inhomogeneous inlets, supersonic combustion, or multiphase combustion. The principal challenges are posed by the uncertainty of the presumed shapes for probability density functions (PDFs) of the flamelet tabulation variables and the coupled process of turbulent diffusion and chemical reaction in turbulent combustion. Recent progress is reviewed from the viewpoint of the reaction manifold, with connections made to other combustion models, as well as the determination of joint (or conditional) PDFs for flamelet manifold parameters (e.g., progress variable, scalar dissipation rates, etc.). Promising improvements have been outlined in computational efficiency and the accuracy of predicted variable fields in simulating complex combustion systems (such as turbulent inhomogeneous combustion, combustion with multi-regime modes, and two-phase combustion). Advances in computational resources, direct numerical simulation data, artificial intelligence, stochastic simulation methods, and other dimension-reduction combustion models will contribute to the development of more accurate and efficient flamelet-like models for engineering applications.

Quantitative studies of NO emissions from various reaction pathways in swirling combustion of hydrogen-enriched methane

Hydrogen-enriched methane is a promising alternative fuel for carbon dioxide reduction. Large eddy simulations of turbulent combustion and NO emission are carried out using a new numerical method, which incorporates an improved NO prediction model and a recently-developed NO reaction pathway separation approach. Quantitative analyses of NO formations from all reaction pathways are conducted in an open swirl burner (combustion in the open environment) with hydrogen-enriched methane at a constant thermal power of 10.85 kW at the atmospheric pressure. Results indicate that different NO reaction pathways response very differently with hydrogen addition. In the open burner with hydrogen enrichment up to 60 %, the prompt NO makes the largest (around 50 %) contribution to the total NO emission, but its relative contribution decreases with hydrogen addition. The thermal NO strongly increases with hydrogen addition, but it accounts for only around 25 % of the total NO emission in hydrogen-enriched cases, because of the reduced high-temperature zone by co-flow air dilution. NO formation from the NNH pathway largely increases and contributes around 20 % to the total NO emission with hydrogen enrichment. Results would provide quantitative understanding of NO emission in the bluff-body stabilized swirling combustion of hydrogen-enriched methane in an open burner.

Study on the application of artificial neural network-based flamelet/progress variable model in supersonic combustion

The flamelet model has the characteristics of high efficiency and physical intuition and has excellent application prospects in supersonic turbulent combustion simulation. Expanding the dimensions of the flamelet model is a potential direction for model development in order to improve its applicability and accuracy, but the accompanying surge in memory is a problem that must be avoided. Therefore, the idea of using the artificial neural network (ANN) model to replace the flamelet database is a feasible exploration currently and has been preliminarily applied in 2D flamelet databases based on central processing unit frameworks. Based on the 3D flamelet database of the flamelet/progress variable (FPV) model, this article studies the strategy of using ANN to replace the flamelet database of the FPV model in a graphics processing unit framework. Due to the significant influence of the progress variable source term and heat release rate on the combustion calculation and the large range of these two parameters, four data processing methods are used to train the parameters separately, and three indicators are used to evaluate the training performance. Subsequently, based on the ANN model using different data processing methods mentioned earlier, calculations are conducted on a hydrogen-fueled supersonic combustion, and the computational accuracy is evaluated. The results indicate that the strategy proposed in this study can screen out artificial neural network replacement models with the same accuracy as the traditional flamelet model.

On the subgrid dissipation concept for large eddy simulation of turbulent combustion

A new subgrid turbulent combustion model for LES is derived based on the eddy dissipation concept (EDC). To be consistent with the LES framework, the relationship between the subgrid quantities and the fine structure ones is built from the unresolved part of the energy cascade, hence the subgrid dissipation concept (SDC). The fine structure quantities are directly computed using the subgrid scale ones. Resolving local laminar flame structures is allowed if the grid resolution permits. The model is examined using premixed piloted jet flames and non-premixed swirl flame. Satisfactory predictions from the SDC model are obtained in the test cases, showing improvements with respect to the standard EDC model. The new SDC model is relatively insensitive to the grid resolution.

Multi-target active subspaces generated using a neural network for computationally efficient turbulent combustion kinetic uncertainty quantification in the flamelet regime

Propagating uncertainties in kinetic models through combustion simulations can provide important metrics on the reliability and accuracy of a model, but remains a challenging and numerically expensive problem especially for large kinetic models and expensive turbulent combustion simulations. Various surrogate model and dimension reduction techniques have previously been applied in order to reduce the cost of forward uncertainty propagation in combustion simulations, but these are often limited to low-dimensional, simple combustion cases with scalar solution targets. In the current work, we developed a neural network-accelerated framework for identifying a low-dimensional active kinetic subspace that applies to the entire temperature solution space of a flamelet table and can capture the mixture fraction and strain rate dependent effects of the kinetic uncertainty. We then demonstrated the computational savings enabled by this framework through a proof-of-concept, flamelet-based application in a Reynolds-averaged Sandia Flame D simulation using a chemical model for methane combustion with 217 reactions. By leveraging the large dimensional compression and low-cost scaling of the active subspace method, offloading the initial dimension reduction gradient sampling onto the laminar flamelet simulations, and accelerating the gradient sampling process with a specifically designed neural network, we were able to estimate the temperature uncertainty profiles across the solution space of the turbulent flame with strong accuracy of 70−85% using just seven perturbed solutions. Additionally, as it occurs entirely within the flamelet table, the cost of identifying the reduced subspace does not scale with the cost of the turbulent combustion model, which is a promising feature of this framework for future application to larger-scale and more complex turbulent combustion applications.

Probabilistic deep learning of turbulent premixed combustion

A probabilistic data-driven approach that models the filtered reaction rate in large-eddy simulation (LES) is investigated. We propose a novel framework that incorporates a conditional generative adversarial network and a Gaussian mixture model to take into account the statistical fluctuations that are present in LES of turbulent reacting flows due to non-resolved subgrid structures, which cannot be predicted by purely deterministic models and machine learning algorithms. The data from a direct numerical simulation of turbulent premixed combustion are spatially filtered using a wide range of filter widths and employed for the training. We extract physically relevant parameters from the database and reduce the input features to the network to the most influential ones based on the result of feature importance analysis. The trained model is then tested on unseen timesteps and untrained LES filter widths, where it is able to accurately predict the distribution of the filtered reaction rate.

Simulation of turbulent non-premixed combustion in pulverized coal from Kalimantan Indonesia

A computational simulation was created to investigate pollutants during coal combustion in a Drop Tube Furnace using Kalimantan coal. Previous research has explored Drop Tube Furnace combustion with Kalimantan coal, but lacked an understanding of combustion phenomena and pollutants, which are challenging to observe experimentally. This study utilized three samples of Kalimantan coal, namely RP, MB, and KC, acquired from various mining sources. This research is new study of simulation combustion coal in Drop Tube Furnace using non-premixed turbulent combustion and the Probability Density Function model with structured grid. The study reports on the temperatures and mass fractions of pollutants, including NOX, SO2, and CO2, along the centerline of the domain. The findings show that RP coal produced the highest combustion temperature, while KC coal produced the lowest. MB coal had the highest CO2 mass fraction, KC coal had the highest NOx value, and RP coal had the highest SO2 value.

Zero-Dimensional Ignition Model of Boron Agglomerates.

Boron primarily exists in the form of agglomerates in ramjet combustion chambers. However, the model used to predict the ignition time of boron agglomerates is usually based on the single-particle assumption, resulting in inaccurate predictions. This study aims to develop a numerical model that can accurately describe the ignition of boron agglomerates. The model is based on the ignition model of a single particle boron proposed by the group of Kuo. Thiele modulus and effectiveness factor are introduced to represent the diffusion resistance of reaction gases in the pores of boron agglomerates. The model includes the necessary physical processes to accurately predict the ignition time. The rates of evaporation and heterogeneous reactions involved in the oxide layer removal process are corrected based on the fact that the diffusion rate of (BO)n in the liquid oxide layer equals to its consumption rate at the oxide-air interface. To evaluate the accuracy of the model, the obtained results for ignition time are compared with experimental data, showing reasonable consistency between them. The model is then applied to investigate the ignition characteristics of boron agglomerates. Parameters, such as initial average pore diameter, oxide layer thicknesses, initial particle diameter, O2 concentration, H2O concentration, and environmental pressure, are studied for their effects on the ignition time. In summary, the boron ignition model established in this study is a powerful tool to investigate the ignition mechanisms and characteristics of boron agglomerates. It can be further coupled with flow analysis for the detailed simulation of turbulent combustion in ramjet combustors.

Quenching modes of local flame–wall interaction for turbulent premixed methane combustion in a constant volume vessel

The quenching mode of local flame–wall interaction (FWI) is investigated for its response to different levels of turbulence intensity as well as its effect on quenching distance, wall heat flux, and near-wall reaction. For that, direct numerical simulations of turbulent premixed methane combustion in a constant volume vessel are carried with initial Karlovitz numbers (Ka) of 1.0, 10.0, and 30.0. Local flame–wall quenching positions are identified based on the local fuel consumption speed during the turbulent combustion process, and the local FWI events have been classified into four quenching modes according to the flame–wall geometric relationships of quenching positions, namely head-on quenching (HOQ), oblique-wall quenching, side-wall quenching (SWQ), and back-on quenching (BOQ). The results show that in the case with higher initial Ka, the flame surface shows a more complicated wrinkled structure due to the flame–turbulence interaction. Meanwhile, the local quenching distance defined based on the identified quenching position is strongly influenced by the near-wall flow, and the range of the local quenching mode extends further to BOQ. However, for all three cases, HOQ and near-HOQ modes account for the majority of local FWI. Wall heat flux and heat release rate (HRR) of near-wall reaction yield high values for the FWI under HOQ or BOQ and are low for SWQ. In addition, there is a discrepancy in the near-wall transportation of some species under different quenching modes, which further leads to the difference in FWI-induced near-wall reaction regarding its total and elementary HRR.

On the effects of adding syngas to an ammonia-MILD combustion regime—A computational study of the reaction zone structure

This study examines the reaction zone structure of syngas/NH3/CH4 blends in moderate or intense low oxygen dilution (MILD) conditions to determine the effects of variations in NH3 and syngas mole fractions upon the stability of the reacting system and NOx emission. A series of axisymmetric turbulent combustion simulations are performed using OpenFOAM and selected reaction mechanisms and, a modified version of the eddy dissipation concept. An increasing amount of syngas is added to the reacting mixtures of NH3/CH4 in MILD conditions. It is observed that a reduction in the NH3 mole fraction increases the OH radical concentration, which implies a reduction in the ignition delay. Further, for a given NH3 mole fraction, an increase in the syngas content of the fuel blend along with a reduction in CH4 leads to a rise in the process maximum temperature near the entrance zone. This reduces the flow temperature gradient, decreases the CH2O mass fraction, and boosts the stability of MILD combustion. Additionally, adding syngas to ammonia for MILD combustion decreases the reaction zone weakening, which is a major problem in MILD combustion of ammonia. In particular, blending syngas with NH3 and reducing the amount of CH4 in the fuel blend increases flame speed and heat generation. When syngas is added to ammonia MILD combustion, NOx and NxO decrease, and more complete combustion is achieved.

Kinetics parameter optimization of hydrocarbon fuels via neural ordinary differential equations

Chemical kinetics mechanisms are essential for understanding, analyzing, and simulating complex combustion phenomena. In this study, a neural ordinary differential equation (Neural ODE) framework is employed to optimize the kinetics parameters of reaction mechanisms. Given experimental or high-cost simulated observations as training data, the proposed algorithm can optimally recover the hidden characteristics in the data. Different datasets of various sizes, types, and noise levels are systematically tested. A classic toy problem of stiff Robertson ODE is first used to demonstrate the learning capability, efficiency, and robustness of the Neural ODE approach. A 41-species, 232-reactions JP-10 skeletal mechanism and a 34-species, 121-reactions n-heptane skeletal mechanism are then optimized with species' temporal profiles and ignition delay times, respectively. Results show that the proposed algorithm can optimize stiff chemical models with sufficient accuracy, efficiency and robustness. It is noted that the trained mechanism not only fits the data perfectly but also retains its physical interpretability, which can be further integrated and validated in practical turbulent combustion simulations. In addition, as demonstrated with the stiff Robertson problem, it is promising to adopt Bayesian inference techniques with Neural ODE to estimate the kinetics parameter uncertainties from experimental data.

Turbulent Flame Propagation in Hydrogen-Air and Methane-Air Mixtures in the Field of Synthetic Turbulence: Direct Numerical Simulation

A technique alternative to the direct numerical simulation of turbulent combustion of gas mixtures is proposed. It is based on the solution of the three-dimensional transport equations for species concentrations and the energy conservation equation in the “synthetic” field of constant-pressure homogeneous, isotropic and statistically stationary (forced) turbulence using the detailed reaction mechanism. The synthetic turbulence with given spatial and temporal correlation functions is generated using the Monte Carlo method, assuming that the components of the vector of fluctuation velocity obey the normal Gaussian distribution. The technique is applied to the problem of turbulent combustion of fuel-lean and stoichiometric mixtures of hydrogen and methane with air at a turbulence intensity up to 10 m/s. The calculated turbulent flame propagation velocities agree satisfactorily with the values measured in the fan-stirred bomb. The predicted volume fractions of active reaction centers H, O, and OH in a turbulent flame are shown to be less than in a laminar flame up to an order of magnitude, which also agrees with the experiment. In general, calculations indicate that the “wrinkled flame” model is applicable to fuel-lean and stoichiometric mixtures of hydrogen and methane with air at turbulence intensities up to 10 m/s

Hybrid Unsupervised Cluster-wise Regression Approach for Representing the Flamelet Tables

A novel hybrid unsupervised cluster-wise regression approach is developed to represent the flamelet tables and speed up the computations in turbulent combustion simulations. The proposed machine learning method utilizes a cluster-wise regression technique where specialized deep neural networks are trained on different parts of the input space. A new recursive unsupervised clustering (RUC) approach is proposed that leads to a lower root-mean-square error by over 30% and improves the individual cluster’s mean absolute error by over 20%. The RUC approach enables identifying different combustion manifolds within the input space. The regression models unique to these identified manifolds accurately capture the thermo-chemical scalars for the given input data set. To balance the tradeoffs between the accuracy of flamelet table predictions and the complexities of the RUC algorithm, a hybrid model is introduced that switches between a single neural network (SNN) which is a single regression model, and the SUC RUC approaches. The hybrid unsupervised cluster-wise-regression model benefits from the lower computational costs of SNN for parts of the input space where the SNN prediction error is low and switches to the unsupervised clustering models, that is, standard (SUC) and recursive (RUC) in areas where the SNN performs poorly. The hybrid model is trained on four-dimensional steady flamelet tables and used to model methane combustion by adopting the Sandia Flame D configuration. The hybrid clustering method more accurately captures the experimental measurements of temperature, carbon monoxide (CO), and hydroxyl radical (OH) mass fraction compared to the SNN model. A new hybrid model offers significant improvements in accuracy as compared to the SNN model along with reduction of computational storage by 2 orders of magnitude compared to the original flamelet approach and reduces the total computational time in 3D reacting flow simulations.

Performance Assessment of Chemical Kinetics Neural Ordinary Differential Equations in Pairwise Mixing Stirred Reactor

Abstract The present study aims to assess the potential of the neural ordinary differential equations (NODE) network for reliable and computationally efficient implementation of chemistry in combustion simulations. Investigations are performed using a hydrogen-air pairwise mixing stirred reactor (PMSR). The PMSR is a zero-dimensional case affordable to study combustion chemistry entailing a similar numerical solution procedure as probability density function methods for turbulent combustion simulations. A systematic approach is presented to apply the NODE, solely trained on canonical constant pressure homogeneous reactor data, to predict complex chemistry and mixing interactions in PMSR. The reactor involves combustion of hydrogen in air described by a finite-rate mechanism with 9 chemical species and 21 reaction steps. The NODE network is shown to accurately capture the evolution of thermochemical variables for different mixing and chemical timescales. It also exhibits a significant reduction in numerical stiffness resulting in improving the computational efficiency and enabling the use of explicit solvers for the integration of chemical kinetics. The assessment results based on PMSR show that compared to direct integration of detailed kinetics, the NODE can achieve significant computational time speedup for a comparable accuracy.