SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Summary of the challenge, approach, results, and impact of co-optimized machine-learned manifold models for research of large eddy simulation of turbulent combustion.
A unique flame-vortex interaction experiment is being operated in microgravity in order to obtain fundamental data to assess the Theory of Flame Stretch which will be used to improve models of turbulent combustion. The experiment provides visual images of the physical process by which an individual eddy in a turbulent flow increases the flame surface area, changes the local flame propagation speed, and can extinguish the reaction. The high quality microgravity images provide benchmark data that are free from buoyancy effects. Results are used to assess Direct Numerical Simulations of Dr. K. Kailasanath at NRL, which were run for the same conditions.
Flow fields involving turbulent flames in premixed gases under a variety of conditions are modeled by the use of a numerical technique based on the random vortex method to solve the Navier-Stokes equations and a flame propagation algorithm to trace the motion of the front and implement the Huygens principle, both due to Chorin. A successive over-relaxation hybrid method is applied to solve the Euler equation for flows in an arbitrarily shaped domain. The method of images, conformal transformation, and the integral-equation technique are also used to treat flows in special cases, according to their particular requirements. Salient features of turbulent flame propagation in premixed gases are interpreted by relating them to the aerodynamic properties of the flow field. Included among them is the well-known cellular structure of flames stabilized by bluff bodies, as well as the formation of the characteristic tulip shape of flames propagating in ducts. In its rudimentary form, the mechanism of propagation of a turbulent flame is shown to consist of: (1) rotary motion of eddies at the flame front, (2) self-advancement of the front at an appropriate normal burning speed, and (3) dynamic effects of expansion due to exothermicity of the combustion reaction. An idealized model is used to illustrate these fundamental mechanisms and to investigate basic aerodynamic features of flames in premixed gases. The case of a confined flame stabilized behind a rearward-facing step is given particular care and attention. Solutions are shown to be in satisfactory agreement with experimental results, especially with respect to global properties such as the average velocity profiles and reattachment length.
This paper describes the time-filtered Navier-Stokes (TFNS) approach capable of capturing unsteady flow structures important for turbulent mixing in the combustion chamber and two different subgrid models used to emulate the major processes occurring in the turbulence-chemistry interaction. These two subgrid models are termed as LEM-like model and EUPDF-like model (Eulerian probability density function), respectively. Two-phase turbulent combustion in a single-element lean-direct-injection (LDI) combustor is calculated by employing the TFNS/LEM-like approach as well as the TFNS/EUPDF-like approach. Results obtained from the TFNS approach employing these two different subgrid models are compared with each other, along with the experimental data, followed by more detailed comparison between the results of an updated calculation using the TFNS/LEM-like model and the experimental data.
Lean combustion in an internal combustion engine is a promising strategy to increase thermal efficiency by leveraging a more favorable specific heat ratio of the fresh mixture and simultaneously suppressing the heat losses to the cylinder wall. However, unstable ignition events and slow flame propagation at fuel-lean condition lead to high cycle-to-cycle variability and hence limit the high-efficiency engine operating range. Pre-chamber ignition is considered an effective concept to extend the lean operating limit, by providing spatially distributed ignition with multiple turbulent flame-jets and enabling faster combustion rate compared to the conventional spark ignition approach. From a numerical modeling perspective, to date, still the science base and available simulation tools are inadequate for understanding and predicting the combustion processes in pre-chamber ignited engines. In this paper, conceptually different RANS combustion models widely adopted in the engine modeling community were used to simulate the ignition and combustion processes in a medium-duty natural gas engine with a pre-chamber spark-ignition system. A flamelet-based turbulent combustion model, i.e., G-equation, and a multi-zone well-stirred reactor model were employed for the multi-dimensional study. Simulation results were compared with experimental data in terms of in-cylinder pressure and heat release rate. Finally, the analysis of the performance of the two models is carried out to highlight the strengths and limitations of the two formulations respectively.
Lean combustion is a promising strategy to increase thermal efficiency in an internal combustion engine, by exploiting a favorable specific heat ratio of the fresh mixture while simultaneously suppressing the heat losses to the cylinder wall. However, unstable ignition and slow flame propagation at fuel-lean conditions lead to large cycle-to-cycle variability and limit the high-efficiency engine operating range. Prechamber ignition is considered an effective concept to extend the lean operating limit, by providing spatially distributed ignition with multiple turbulent flame-jets and enabling a faster combustion rate compared to the conventional spark ignition approach. From a numerical modeling standpoint to date science base and available simulation tools are inadequate to properly understand and predict the combustion processes in prechamber ignited engines. In this paper, conceptually different Reynolds-averaged Navier–Stokes (RANS) combustion models widely adopted in the engine modeling community are used to simulate the ignition and combustion processes in a medium-duty natural gas engine with a prechamber spark-ignition system. Here, a flamelet-based turbulent combustion model, i.e., G-equation, and a multizone well-stirred reactor model are employed for this modeling study. Simulation results are compared with experimental data in terms of in-cylinder pressure and heat release rate. Finally, the analysis of the performance of the two models is carried out to highlight the strengths and limitations of the two evaluated approaches.
Combustion of premixed reactants in a turbulent flow is a classical but unresolved problem. The key problem is to explain the following data: the maximal turbulent and laminar burning velocities u(sub t) and u(sub L) occur at different equivalence ratios Phi. It is known that the equivalence ratio varies along a curved flame if molecular diffusivity D(sub fuel) does not equal D(sub oxygen). However, the mean flame radius of curvature is much larger than the laminar flame thickness delta-L. Therefore, significant influence of preferential diffusion should occur only if the flame propagation speed varies with flame curvature. This conclusion agrees with Zel'dovich's long-standing idea about the important role of leading points of a flame. The main objective of this paper is to prove Zel'dovich's hypothesis. An equation for the mean flame surface area density (MFSAD) is employed for this purpose. The second objective of this paper is to suggest a different approach to the derivation of the equation for MFSAD. It is based on the pdf equation for the reaction progress variable C and the relation between the pdf and MFSAD. This treatment suggests an entirely different closure assumption.
Here, this study presents a novel reduced-order modeling framework, Time-Dependent Bases with Local CUR decomposition (TDB-L-CUR), designed to efficiently and accurately approximate the species transport equations in reacting flow simulations. The method extends the existing TDB-CUR approach for chemically reacting flows (Jung et al. Comput. Methods Appl. Mech. Engrg. 437 (2025) 117758), which leverages matrix decomposition techniques to form a global-in-space, time-dependent low-dimensional manifold. While TDB-CUR performs well in homogeneous systems, it may be less well-suited to spatially heterogeneous systems such as turbulent flames, where higher-rank approximations are typically required. The proposed TDB-L-CUR framework introduces two methodological extensions to the baseline approach. First, it applies unsupervised clustering to partition the physical domain into distinct regions, enabling spatially localized manifold construction, thereby reducing the rank required for the reduced-order representation. Second, it incorporates a computational singular perturbation (CSP)-based scheme for identifying and penalizing fast species, allowing for spatio-temporally adaptive mitigation of chemical stiffness. The proposed framework is validated on a hierarchy of test cases, including a one-dimensional premixed flame, a two-dimensional nonpremixed ignition case with vortex interaction, and a three-dimensional turbulent premixed flame. TDB-L-CUR significantly improves accuracy over TDB-CUR while further reducing computational cost. The fully on-the-fly formulation of TDB-L-CUR (i.e., requiring no offline training or prior knowledge) makes it a robust and scalable tool for reduced-order modeling of reactive flows.
Reduced-order manifold combustion models are commonly used to lower the cost of reacting Large Eddy Simulations (LES) and may be derived either from physical principles as in flamelet models or from data-driven methods like Principal Component Analysis (PCA). In either case, neural networks are increasingly used as part of these models to provide a nonlinear mapping between a small set of pre-defined variables that parameterize the manifold and outputs of interest, such as reaction rates. In this work, we propose a new manifold-based modeling approach that combines the definition of the manifold-parameterizing variables (linear combinations of species), the nonlinear mapping to the outputs, and closure of filtered quantities for LES into the structure of a single neural network. This allows the process used to train the neural network to simultaneously optimize both the functional form of the model and the identities of the inputs to the model. The new approach can flexibly incorporate thermochemical data from any combustion system; if trained on data from 1D flames it can be interpreted as an optimized flamelet model, but it can also be used to learn models from data from more complex configurations. This work presents a priori evaluations of the new approach in both contexts. Evaluation using data from 1D premixed flames demonstrates the physical interpretability of the manifold variables generated by the new approach. Evaluation using data from direct numerical simulations of turbulent flames shows improved predictions relative to either flamelet or PCA-based models in a more complex configuration.
The work in numerical modeling is focused on the use of the random vortex method to treat turbulent flow fields associated with combustion while flame fronts are considered as interfaces between reactants and products, propagating with the flow and at the same time advancing in the direction normal to themselves at a prescribed burning speed. The latter is associated with the generation of specific volume (the flame front acting, in effect, as the locus of volumetric sources) to account for the expansion of the flow field due to the exothermicity of the combustion process. The model was applied to the flow in a channel equipped with a rearward facing step. The results obtained revealed the mechanism of the formation of large scale turbulent structure in the wake of the step, while it showed the flame to stabilize on the outer edges of these eddies.
A computationally viable scheme to account for chemical reaction in turbulent flows is presented. The multivariate beta-pdf model for multiple scalar mixing forms the basis of this scheme. Using the model scalar joint pdf and a general form of the instantaneous reaction-rate, the unclosed chemical reaction terms are expressed as simple functions of scalar means and the turbulent scalar energy. The calculation procedure requires that the mean scalar equations and only one other transport equation - for the turbulent scalar energy - be solved.
A two dimensional combustion tunnel was constructed to study a lean premixed turbulent propane/air flame stablized behind a rearward facing step. Studied were: (1) the existence and importance of large coherent structures in turbulent reacting and nonreacting free shear layers behind the steps; (2) the effect of inlet temperature and reference velocity on combustion efficiency; (3) CO, NO2 and NO sub x production in the flame; and (4) the blowout and upstream propagation of the flame. In the ranges studied, the large coherent structures dominated both the reacting and the nonreacting free shear layers behind the step. The growth of the vortices and the propagation of the flamer were intimately linked. Vortex pairing was observed to be one of the mechanisms for introduction of fresh reactants into the shear layer and growth of the shear layer. Probe composition measurements of the flame showed that, in the recirculation zone, the reaction was above 99 percent complete, CO and unburnt hydrocarbons were above the equilibrium level NO sub x concentration was far below the equilibrium level and NO2 comprised a negligible fraction of NO sub x.
In premixed turbulent flames the presence of intense mixing zones located in front of and behind the flame surface leads to a requirement to study the behavior of iso-concentration surfaces defined for all values of the progress variable (equal to unity in burnt gases and to zero in fresh mixtures). To support this study, some theoretical and mathematical tools devoted to level surfaces are first developed. Then a database of direct numerical simulations of turbulent premixed flames is generated and used to investigate the internal structure of the flame brush, and a new pdf model based on the properties of iso-surfaces is proposed.
A complete second-order closure program for the investigation of chemically reacting turbulent flowfields is being developed. Many of the turbulence models being used have been developed and tested during previous studies of a variety of incompressible and compressible shear flows. A number of additional scalar correlations have to be modeled for chemically reacting flows, and a 'typical eddy' model has been developed for the joint probability density function for all the scalars. A simpler model in which higher-order scalar correlations are set zero is also being used. The program has been recently extended to multi-species, multi-step chemical reactions. Computations have been performed using the second-order closure program to study the turbulence-chemistry interaction in DF chemical laser flows and in a hydrogen-air diffusion flame. The results demonstrate the importance of the inclusion of the concentration and temperature fluctuation correlations in turbulent reacting flows.
In a recent attempt by the authors at predicting measurements in coaxial supersonic turbulent reacting mixing layers involving H2 and air, a number of discrepancies involving the concentrations and their variances were noted. The turbulence model employed was a one-equation model based on the turbulent kinetic energy. This required the specification of a length scale. In an attempt at detecting the cause of the discrepancy, a coupled k-omega joint probability density function (PDF) is employed in conjunction with a Navier-Stokes solver. The results show that improvements resulting from a k-omega model are quite modest.
The application of pulsed resonance spectroscopy to the measurement of species concentrations in chemically reacting turbulent flows is discussed. The theory of such measurements is developed. An uncertainty analysis is performed, and the results are applied to typical combustion conditions. Experimental aspects are discussed. The method shows a significant theoretical improvement in accuracy over Raman and Reyleigh scattering techniques, and no significant barriers appear to block its development.
Manifold-based models offer a computationally efficient alternative to directly transporting the thermochemical state in computational simulations of turbulent reacting flows, projecting the high-dimensional thermochemical state-space onto a low-dimensional manifold. Recent efforts have yielded a manifold-based model applicable to multi-modal combustion, enabling reconstruction of the thermochemical state from solutions to two-dimensional manifold equations in mixture fraction and generalized progress variable that are parameterised by three scalar dissipation rates. In coarse-grained simulations such as Large Eddy Simulation (LES), closure of the multi-modal manifold equations and subfilter variances/covariance requires closure of three filtered scalar dissipation rates. Here, the present work adopts a data-based approach, providing closure for the three filtered scalar dissipation rates via deep neural networks (DNNs). High-fidelity datasets corresponding to an autoigniting n-dodecane jet flame and a bluff body swirl-stabilized confined lifted spray flame of two aviation fuels (Jet-A and C1) with different ignition propensities are leveraged to generate training data that spans a diverse range of thermodynamic conditions and combustion modes, including low- and high-temperature ignition regimes in addition to premixed and nonpremixed behaviour. A final DNN model is trained to enforce inherent physical constraints by learning nonlinear functional transformations of the three filtered scalar dissipation rates. The generalizability of this constrained DNN model is demonstrated a priori via conditional statistics evaluated on the lifted spray flame with C1–a configuration that had not been included in the training data. Excellent DNN agreement with conditional DNS statistics is observed, and integrated gradients are computed to identify the most sensitive input variables. The similarity of the marginal PDFs of the most informative input variables and outputs across configurations are quantified via the Wasserstein metric, demonstrating that data-based models may successfully generalize to unseen parametric conditions so long as the most informative input variables share similar distributions across training and testing datasets.
A simplified theoretical model for turbulent methane/air flames from a subscale power plant burner is developed and applied to predict the stability of flames as a function of (1) ambient atmosphere temperature and composition, (2) initial preheat temperature, and (3) percentage of external product gas recirculation. The flame model equates the rate at which mass is supplied per unit volume to the rate it is chemically reacted per unit volume using a one-step reaction and a perfectly stirred reactor (PSR). The model contains a free parameter which is evaluated by comparing the theoretical results to experimental results available in the literature. The stability results show that premixed, turbulent flames are stable over a wide equivalence ratio range if the ambient gas is air; however, if combustion products are entrained, the stability is sharply reduced.