This study explores the use of machine learning (ML) techniques, specifically artificial neural networks (ANN) and gradient boosting(GB), to replace conventional flamelet look-up tables in turbulent combustion modeling. The approach is based on the unsteadyflamelet/progress variable (UFPV) formulation, which serves as the foundation for the underlying flamelet manifolds. A central focusis placed on the transformation of high-dimensional tabulated chemistry data into efficient ML surrogates suitable for integrationinto computational fluid dynamics (CFD) simulations. Accordingly, various data pre-processing strategies and training methodologiesare evaluated, and the predictive performance of the resulting ML models is thoroughly assessed. To demonstrate the generalityand robustness of the proposed framework, the methodology is applied to multiple fuels, including n-dodecane (C12H26), n-heptane(C7H16), and oxymethylene ether (OME34). Beyond predictive accuracy, the study emphasizes key practical advantages of theML-based approach, namely reduced memory footprint and lower computational overhead by eliminating runtime interpolation,quantifying its potential to enhance the efficiency of combustion simulations.

The present study investigates the formulation of surrogates for Jet-A and sustainable aviation fuel (SAF) by means of a chemical reactor network (CRN) to predict emissions in gas-turbine combustors. The modeling framework is used to analyze the effects of fuel composition, specifically the replacement of Jet-A with SAF as well as the substitution of aromatics in Jet-A with cycloalkanes. This approach enables the examination of the effects of the fuel class and molecular structure independently of global exhaust emission trends. A comprehensive chemical kinetic mechanism incorporating 8478 species and 33,318 reactions was used to model jet fuel surrogates. This mechanism, validated against ignition delay times, laminar flame speeds, and extinction strain rates, accurately predicted combustion characteristics for iso-cetane and iso-dodecane, key components of SAF. Surrogate fuels for Jet-A and alternative fuels were formulated targeting critical properties such as density and cetane number. Surrogate simulation results show strong alignment with experimental data on ignition delay and flame speed, further confirming the reliability of the surrogates. The CRN model was developed using the CFM56 engine data and validated against the International Civil Aviation Organization emission benchmarks. After that, two different fuel replacement scenarios, involving Jet-A, are investigated. In the first one, Jet-A is compared to 100% SAF, which helps assess expected differences in combustion performance if a fully renewable fueling supply is followed. Results show that NO x emissions are unaffected by SAF, aligning with previous experimental studies. Polycyclic aromatic hydrocarbons (PAH) decrease by 93% for SAF compared to Jet-A, while Jet-A produces more CO due to its aromatic content. Substituting aromatics with cycloalkanes in Jet-A reduces PAH emissions by up to 96 or 92%, depending on whether all aromatics are replaced or only diaromatics. In a second scenario, aromatics are removed from conventional Jet-A and replaced with cycloalkane species, which can still hold the swelling properties needed by the fueling system. In terms of combustion, cycloalkane substitution leads to slightly increased CO emissions and reduced flame temperatures. These results demonstrate the potential of SAFs and cycloalkanes in reducing soot precursors while maintaining a highly similar performance in the combustion process. Overall, the proposed CRN framework provides a good example of how a predictive and computationally efficient tool can help in the early evaluation of alternative aviation fuels under realistic gas-turbine combustor conditions.

Accurately controlling combustion and achieving cleaner emissions in compression ignition engines need a better understanding of soot evolution. In the present study, the transient characteristics of in-flame soot evolution of spray A within the engine combustion network are numerically investigated under high-temperature and high-pressure conditions using a parcel tracing methodology. A virtual parcel driven by the current flow field, recording the local information to understand the evolution of soot. A two-equation soot model has been implemented based on the OpenFOAM platform, coupling with a ∑−Y Eulerian spray model, and unsteady flamelet progress variable combustion model. The computed results indicate good agreement with the experimental ones under studied conditions. Results show that oxygen concentration significantly affects the onset soot location, which occurs at some radial distance away from the spray axis under low oxygen conditions, while it aligns along the spray axis for high oxygenconditions. The peak soot location of quasi-steady flames is concentrated around an equivalence ratio region (2<Φ<2.2) regardless of operating conditions. The chemical source of soot production is dominated by surface growth rate after the traced parcels pass through flame liftoff length where the local temperature exceeds 1500 K.

This work is focused on the modeling and analysis of soot formation and oxidation in the pressurized ethylene-based model burner investigated at DLR. This burner features a dual swirler configuration for the primary air supply and includes secondary dilution jets inside the combustion chamber, showing reacting flow characteristics representative of the RQL combustor technology. Large-eddy simulations (LES) of the DLR burner are conduced here to assess a coupling approach between flamelet generated manifold (FGM) chemistry and discrete sectional method (DSM) based soot model with clustering method. First, a validation of the numerical results is conducted for the gas velocity and temperature fields, and good agreement is obtained for both mean and fluctuating quantities. The Soot Volume Fraction (SVF) computed from LES shows a satisfactory agreement with the experimental data in both SVF distribution and magnitude. The analysis also includes a numerical investigation of the soot production and the Particle Size Distributions (PSD). Finally, the configuration without secondary air is evaluated and an accurate prediction of the SVF field is also obtained. In this case, the absence of dilution air strongly influences the central region of the combustion chamber, and soot distribution and PSD are mainly affected by transport and dilution, not oxidation. It is finally concluded the proposed modeling framework is capable of predicting the soot field and particle size distributions inside the combustor for both operating conditions.

Oxymethylene dimethyl ethers (OMEx), having a chemical formula of CH3O-(CH2O)(x)-CH3 where x varies from 1 to 5, have been widely considered as a promising fuel to partially replace diesel in CI engines in terms of reducing soot emissions. This work is focused on developing a reduced primary reference fuel (PRF)-OMEx chemical mechanism to better describe the combustion and emission characteristics of gasoline/diesel blends with OMEx. The novelty of this work lies in the fact that the OMEx part of the mechanism is represented not only by OME3 as done in most studies found in literature, but also with other OME chain lengths that is, OME2-4 which are considered to be optimum and better represent the commercial OMEx blends. For this purpose, a detailed OMEx mechanism is reduced by applying different reduction techniques considering a wide range of operating conditions including pressure, temperatures, equivalence ratios and fuel compositions. The result is merged with an already validated PRF mechanism to form a reduced PRF-OMEx mechanism consisting of 213 species and 840 reactions. The newly formed mechanism is validated against a wide set of experimental data including ignition delay times, laminar flame speeds and species concentration profiles. Furthermore, a rigorous set of numerical simulations for various diesel-OMEx blends in a compression ignition engine are carried out at two different operating points to validate the developed mechanism. Simulation results highlight that the developed mechanism not only replicates the experimental behavior in terms of in-cylinder pressure and heat release rate but exhibits a better combustion phasing closer to experimental data when compared with other mechanisms where only OME3 is utilized to represent OMEx. Overall, the developed PRF-OMEx mechanism proves to be realistic and suitable for application in engine combustion simulations involving gasoline/diesel and OMEx blends.

The present work focuses on the derivation and evaluation of a chemical kinetic mechanism of primary reference fuel [(PRF, binary blends of n-heptane and isooctane)] with a homogeneous reactors approach starting from a detailed one. Results show that the optimized mechanism can replicate the results of the detailed one with high accuracy. The mechanism is integrated into a computational fluid dynamics workflow combining a Reynolds-averaged Navier-Stokes approach, a diffuse-interface spray, and an unsteady flamelet progress variable combustion model. The workflow is validated against spray combustion measurements following the standards of the engine combustion network (ECN). Test cases sweep binary blends of PRF fuels from pure n-heptane to pure iso-octane using an ECN Spray A nozzle. The model can provide accurate predictions of typical reacting spray metrics, such as ignition delay and lift-off length, which have been evaluated following a reconstruction of the experimental methods, namely schlieren and OH* chemiluminescence. Different definitions of the previous combustion metrics have been compared. The model captures the decreasing reactivity with increasing isooctane fraction, which results in flame stabilizing at much leaner conditions. However, deficiencies are observed for low reactivity cases, either with high PRF or low-temperature cases.

Ammonia is seen as a promising fuel to replace fossil fuels internal combustion engines, given the advantages of zero carbon emission and extensive experience in its synthesis at large scale. The ignition of ammonia in direct injection engine is still a challenge due to its unfavorable combustion characteristics, but the investigation of ammonia spray characteristics is an important step to shed light on the optimum combustion strategy. In this paper, the evolution of direct injection ammonia sprays was measured at engine-like conditions generated in a constant pressure facility by means of diffused back-illumination imaging and schlieren. In a first step, ammonia spray behavior is compared against that of Diesel fuel. Experimental results show a slower penetration of ammonia mainly due to a lagging starting penetration phase, which cannot be later recovered. In terms of liquid length, initial stages show the same pattern as for ammonia, while a shorter stabilized liquid length can be observed. Parametric trends over different operating conditions show a similar behavior to Diesel against ambient density and temperature, as well as with injection pressure and nozzle diameter. Such results, backed up by the application of a 1D spray model developed for Diesel-like fuels, suggest that ammonia sprays under non-reacting conditions have a mixing-controlled evolution.

The present work reports an experimental evaluation of the flame stabilization, structure and pollutant emissions for a swirl-stabilized premixed burner at atmospheric conditions. The burner has been designed to accommodate two inlet streams, namely a swirling air and a non-swirling fuel-air mixture, which are fed into a mixing chamber. By modifying the air mass distribution between the swirling and non-swirling streams, the burner can produce different swirl numbers at the chamber, as well as different fuel-air mixture stratification. A central tube acts as a bluff body to help stabilize the flame. Experimental diagnostics include high speed flame visualization in terms of OH* chemiluminescence, and NOx pollutant at the burner outlet. The study has been carried out using both hydrogen and methane as fuels. A sweep of conditions has been carried out to identify the lean blow-off limit over the burner operational map. Results for methane indicate that the lean blow-off limit is strongly affected by the air distribution between the swirling and non-swirling parts. In contrast, air distribution has a more significant effect on NOx emissions for hydrogen. Different flame shapes are analyzed over the operational map, where strong differences in lift-off, flame width, and reaction zone chemiluminescence have been identified.

The solution to the dilemma of carbon footprint of internal combustion engines and pollutant emissions is necessary for the survival of this technology. In this context, alternative fuels have shown great potential in terms of achieving cleaner combustion and compliance with ever increasing pollutant emissions regulations. This work is focused on the study of two promising alternative fuels as Hydrotreated vegetable oil (HVO), which is a biofuel and Dimethoxymethane also termed as OME1, which is an e-fuel. A comprehensive numerical approach has been followed to study these fuels. Primarily a compact reaction mechanism having 121 species and 678 reactions has been developed which can be utilized to perform 3D CFD simulations of blends of these fuels. Secondly, a detailed numerical investigation including combustion and emissions analysis has been carried out. Results show that the developed mechanism is able to offer predictions, which match the experimental behavior observed in various combustion parameters and thus can be utilized for compression ignition engine applications involving these promising fuels. In addition, the numerical analysis also highlights that a reduction of 50% and 37% in terms soot and NOx emissions respectively can be achieved by addition of 30% OME1 in the blend containing HVO, suggesting that these blends can be utilized in unmodified CI engines to break the soot-NOx tradeoff without significantly penalizing the energy loss.

This study presents an assessment of the Partially Stirred Reactor (PaSR) as a subgrid model for large eddy simulations (LES) of turbulent premixed combustion. The PaSR-LES approach uses a skeletal mechanism for methane/air combustion, and requires the transport of all the species, with a closure for the filtered source terms. The rate of progress for each reaction is given by the mixing and chemical time scales, which are computed from global flame parameters and a turbulent time scale respectively. This model is applied to a swirled combustor exhibiting a V-flame shape attached to the nozzle, subjected to heat loss. LES are carried out for two distinct equivalence ratios at atmospheric pressure. The flow fields and the thermochemical states from PaSR-LES are compared with the experimental data and solutions based on Flamelet Generated Manifolds (FGM). The results show good correlation with the experiments and FGM-LES, though also some sensitivity to the resolution. The approach also reproduces well the effect of heat loss, which is determined by the use of a chemical time scale given by a progress variable. Dedicated analysis of the swirl-stabilized flame on different regions is conducted evaluating the capabilities of the model to reproduce the burning velocity, flame shape and flame structure.

Numerical simulations using LES with tabulated chemistry based on the Flamelet Generated Manifold (FGM) method are conducted with the aim to understand the influence of mixture stratification on the flame dynamics on a swirled burner with bluff body. This configuration consists of a main injection of premixed fuel (CH4) with air in rich conditions, injected into a coflow of swirled air which will further mix along a premixing length prior to the combustor, resulting into a global mixture burning in lean conditions. While maintaining the global equivalence ratio inside the combustion chamber to ϕg = 0.8, three operating conditions are simulated, which differ on the mass fraction of air premixed with methane (5, 10 and 15 %). Different flame topologies are obtained, and a qualitative comparison with experiments is performed.

Flamelet models are widely used in computational fluid dynamics to simulate thermochemical processes in turbulent combustion. These models typically employ memory-expensive lookup tables that are predetermined and represent the combustion process to be simulated. Artificial neural networks (ANNs) offer a deep learning approach that can store this tabular data using a small number of network weights, potentially reducing the memory demands of complex simulations by orders of magnitude. However, ANNs with standard training losses often struggle with underrepresented targets in multivariate regression tasks, e.g., when learning minor species mass fractions as part of lookup tables. This paper seeks to improve the accuracy of an ANN when learning multiple species mass fractions of a hydrogen (\ce{H2}) combustion lookup table. We assess a simple, yet effective loss weight adjustment that outperforms the standard mean-squared error optimization and enables accurate learning of all species mass fractions, even of minor species where the standard optimization completely fails. Furthermore, we find that the loss weight adjustment leads to more balanced gradients in the network training, which explains its effectiveness.

The present work focuses on the prediction of pollutant emissions in gas turbine engines by means of a reactor network considering chemical kinetics. The network was developed based on the CFM56 -7B27/B1F aviation engine, calibrated with the pollutant emission data (EINOx, EICO and EIHC) provided by ICAO at maximum power and compared to other operating conditions. A study of local species formation in every reactor was carried out to analyze the differences and similarities between the model and experimental data.

Mitigation of the carbon footprint of internal combustion engines is mandatory to ensure a future for this technology. Within this scope, e-fuels are considered a potential solution to replace conventional fossil fuels. However, in some cases, their physical and chemical properties are so different that its application in conventional engines is complex. For this reason, this work focuses on the study of oxymethylene ethers (OME X ) as a potential low-carbon fuel alternative. The aim is to improve the understanding of the combustion process of these e-fuels when they replace fossil Diesel in internal combustion engines under equivalent operating conditions. To achieve this objective, a computational fluid dynamics model of an optical compression ignition engine has been developed. The operating conditions chosen are representative of a medium load point of the engine, which coincide with experimental work previously done on this platform. n-Heptane was used as surrogate of fossil Diesel while OME X was simulated as a simpler mixture of oxymethylene ether molecules. Results show remarkable differences between Diesel and OME X . This fuel provides lower equivalence ratio fields. Thus, oxidation reactions are promoted in wider areas within the combustion chamber, leading to a faster combustion process. Besides, the soot formation is also drastically decreased in comparison to the other fuel. These results have been corroborated with experimental information.

Poly-Oxymethylene Dimethyl Ethers OMEx are synthetic and potentially-renewable fuels that lead to a notable reduction of the lifecycle CO2 emissions while promoting lower soot emissions than conventional Diesel fuel. In the present contribution, a computational study with a single component OME1 and a multicomponent OMEx fuel has been carried out under reference Spray A conditions from the Engine Combustion Network (ECN), which mimic in-cylinder conditions representative of Diesel engines. For both fuels, three ambient temperature conditions have been swept at constant ambient density. Calculations have been carried out using an Unsteady Flamelet Progress Variable (UFPV) combustion model and detailed chemical mechanisms. For both OMEx-type fuels, low temperature ignition in flamelet configurations start in lean mixtures, which shifts towards the fuel -rich zone and eventually leads to high temperature ignition, similar to typical hydrocarbons. In agreement with corresponding fuel cetane numbers, ignition of OMEx occurs at timings similar to those of n-dodecane, which is the reference fuel for ECN studies, while a delayed ignition is obtained for OME1. However, the actual difference in ignition timing between OMEx and n-dodecane depends on diffusion in the mixture fraction space. Moreover, ignition in spray calculations seems to occur fully on the lean side, especially for OME1, as well as for the low temperature cases. This difference in ignitable mixture range between the canonical flamelet configuration and the spray calculations results from the finite residence time for relevant mixtures in the latter case, compared to an infinite residence time in flamelets. Comparison with experiments show that the modeling approach predicts most combustion metrics for both fuels and temperature values. The combination of ambient temperature and fuel-related reactivity has enabled a transition from a short ignition lifted diffusion flame structure (OMEx at 1000-900 K) towards long ignition cases, where lift-off length may eventually be longer than the maximum length of the stoichiometric surface. This results in a reaction front stabilization at very lean conditions, i.e. a type of lean mixing-controlled flame.

Synthetic fuels will play a major role on the reduction of pollutant emissions and carbon footprint of ICE engines. The use of renewable energy and CO2 for its production maps out a promising route to achieve ICE carbon neutrality. Among these fuels, oxymethylene ethers are also interesting for their potential to drastically reduce soot formation. However, a proper characterization of their properties and behaviour is mandatory to reach full implementation in commercial engines. For this reason, this work presents a detailed characterization of the flame structure of two types of oxymethylene ethers (OMEX and OME1) using high-speed chemiluminescence imaging and Planar Laser-Induced Fluorescence (PLIF). Test were performed in a constant-pressure combustion vessel, with the operating conditions and two different injector nozzles from the Engine Combustion Network (Spray A and Spray D). Regions associated with low-temperature chemical reactions are observed thanks to the formaldehyde PLIF while evolution of the high-temperature reactions has been analysed based on hydroxyl excited state (OH*) chemiluminescence and hydroxyl PLIF. On-resonant and off-resonant measurements were performed for OH PLIF with a dye laser in order to remove other radiation sources not linked to OH fluorescence, whilst 355 nm radiation from the third harmonic of a Nd:YAG laser are used to excite CH2O molecules. On the one hand, the results show that the combustion of OME1 with Spray A is characterized by a flame structure very different to that of a diffusion flame. It is characterized by large cool flame region followed by a short high-temperature zone. On the other hand, the combustion of OMEX with both nozzles and OME1 with Spray D show more similarities with a diffusion flame structure. The stoichiometry of the fuel and the equivalence ratio fields achieved strongly affect the structure and latter evolution of the flames.

Despite recent advances towards powertrain electrification as a solution to mitigate pollutant emissions from road transport, synthetic fuels (especially e- fuels) still have a major role to play in applications where electrification will not be viable in short-medium term. Among e-fuels, oxymethylene ethers are getting serious interest within the scientific community and industry. Dimethoxy methane (OME1) is the smaller molecule among this group, which is of special interest due to its low soot formation. However, its application is still limited mainly due to its low lower heating value. In contrast, other fuel alternatives like hydrogenated vegetable oil (HVO) are considered as drop-in solutions thanks to their very similar properties and molecular composition to that of fossil diesel. However, their pollutant emission improvement is limited. This work proposes the combination of OME1 and HVO as an alternative to fossil diesel, to achieve noticeable soot emission reductions while compensating for the different properties of the first fuel. The aim of this work is to provide insight into the combustion characteristics of blends of these two fuels. For this purpose, experimental and numerical studies are combined. In this context, n-dodecane is proposed as a surrogate for HVO simulation based on the high similarities experimentally observed between both fuels. Then, a compact kinetic mechanism is developed and validated, combining individual OME1 and n-dodecane mechanisms. Results confirm that the numerical approach followed was able to capture the experimental behavior of these blends in terms of heat release rate, in-cylinder pressure and soot formation. An increase of the OME1 content in the blend greatly influences the combustion process. The ignition delay, as well as the premixed combustion phase peak, increase with the OME1 percentage in the blend. However, HVO helps on limiting this effect while remarkable soot formation reductions are still achieved thanks to OME1.