Publications

Abstract not provided.

Direct numerical simulation (DNS) of a high Karlovitz number (Ka) CH4/air stratified premixed jet flame was performed and used to provide insights into fundamentals of turbulent stratified premixed flames and their modelling implications. The flame exhibits significant stratification where the central jet has an equivalence ratio of 0.4, which is surrounded by a pilot flame with an equivalence ratio of 0.9. A reduced chemical mechanism for CH4/air combustion based on GRI-Mech3.0 was used, including 268 elementary reactions and 28 transported species. Over five billion grid points were employed to adequately resolve the turbulence and flame scales. The maximum Ka of the flame in the domain approaches 1400, while the jet Damköhler number (Dajet) is as low as 0.0027. The flame shows early stages of CH4/air combustion in the near field and later stages in the far field; the separation of combustion stages can be largely attributed to the small jet flow timescale and the low Dajet. The gradient of equivalence ratio in the flame normal direction, dϕ/dn, is predominantly negative, and small-scale stratification was found to play an important role in determining the local flame structure. Notably, the flame is thinner, the burning is more intense, and the levels of the radical pools, including OH, O and H, are higher in regions with stronger mixture stratification. The local flame structure is more strained and less curved in these regions. The mean flame structure is considerably influenced by the strong shear, which can be reasonably predicted by unity Lewis number stratified premixed flamelets when the thermochemical conditions of the reactant and product are taken locally from the DNS and the strain rates close to those induced by the mean flow are used in the flamelet calculation. An enhanced secondary reaction zone behind the primary reaction zone was observed in the downstream region, where the temperature is high and the fuel concentration is negligible, consistent with the observed separation of combustion stages. The main reactions involved in the secondary reaction zone, including CO + OH⇔CO2 + H (R94), H + O2 + M⇔HO2 + M (R31), HO2 + OH⇔H2O + O2 (R82) and H2 + OH⇔H2O + H (R79), are related to accumulated intermediate species including CO, H2, H, and OH. The detailed mechanism of intermediate species accumulation was explored and its effect on chemical pathways and heat release rate was highlighted.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Exascale computing promises quantities of data too large to efficiently store and transfer across networks in order to be able to analyze and visualize the results. We investigate compressed sensing (CS) as an in situ method to reduce the size of the data as it is being generated during a large-scale simulation. CS works by sampling the data on the computational cluster within an alternative function space such as wavelet bases and then reconstructing back to the original space on visualization platforms. While much work has gone into exploring CS on structured datasets, such as image data, we investigate its usefulness for point clouds such as unstructured mesh datasets often found in finite element simulations. We sample using a technique that exhibits low coherence with tree wavelets found to be suitable for point clouds. We reconstruct using the stagewise orthogonal matching pursuit algorithm that we improved to facilitate automated use in batch jobs. We analyze the achievable compression ratios and the quality and accuracy of reconstructed results at each compression ratio. In the considered case studies, we are able to achieve compression ratios up to two orders of magnitude with reasonable reconstruction accuracy and minimal visual deterioration in the data. Our results suggest that, compared to other compression techniques, CS is attractive in cases where the compression overhead has to be minimized and where the reconstruction cost is not a significant concern.

Considering that simulations of turbulent combustion are computationally expensive, this chapter takes a decidedly different perspective, that of high-performance computing (HPC). The cost scaling arguments of non-reacting turbulence simulations are revisited and it is shown that the cost scaling for reacting flows is much more stringent for comparable conditions, making parallel computing and HPC indispensable. Hardware abstractions of typical parallel supercomputers are presented which show that for design of an efficient and optimal program, it is essential to exploit both distributed memory parallelism and shared-memory parallelism, i.e. hierarchical parallelism. Principles of efficient programming at various levels of parallelism are illustrated using archetypal code examples. The vast array of numerical methods, particularly schemes for spatial and temporal discretization, are examined in terms of tradeoffs they present from an HPC perspective. Aspects of data analytics that invariably result from large feature-rich data sets generated by combustion simulations are covered briefly.

This supplementary material complements the article and provides additional information to the chemical mechanism used in this work, boundary conditions for the LES con guration and table generation, comparisons of axial velocities, results from a LES/ nite-rate chemistry (FRC) approach, and results from the LES/DTF/SPF approach with a particular chemistry table that is generated using a single strained premixed amelet solution.

In order to achieve exascale systems, application resilience needs to be addressed. Some programming models, such as task-DAG (directed acyclic graphs) architectures, currently embed resilience features whereas traditional SPMD (single program, multiple data) and message-passing models do not. Since a large part of the community's code base follows the latter models, it is still required to take advantage of application characteristics to minimize the overheads of fault tolerance. To that end, this paper explores how recovering from hard process/node failures in a local manner is a natural approach for certain applications to obtain resilience at lower costs in faulty environments. In particular, this paper targets enabling online, semitransparent local recovery for stencil computations on current leadership-class systems as well as presents programming support and scalable runtime mechanisms. Also described and demonstrated in this paper is the effect of failure masking, which allows the effective reduction of impact on total time to solution due to multiple failures. Furthermore, we discuss, implement, and evaluate ghost region expansion and cell-to-rank remapping to increase the probability of failure masking. To conclude, this paper shows the integration of all aforementioned mechanisms with the S3D combustion simulation through an experimental demonstration (using the Titan system) of the ability to tolerate high failure rates (i.e., node failures every five seconds) with low overhead while sustaining performance at large scales. In addition, this demonstration also displays the failure masking probability increase resulting from the combination of both ghost region expansion and cell-to-rank remapping.

Obtaining multi-process hard failure resilience at the application level is a key challenge that must be overcome before the promise of exascale can be fully realized. Previous work has shown that online global recovery can dramatically reduce the overhead of failures when compared to the more traditional approach of terminating the job and restarting it from the last stored checkpoint. If online recovery is performed in a local manner further scalability is enabled, not only due to the intrinsic lower costs of recovering locally, but also due to derived effects when using some application types. In this paper we model one such effect, namely multiple failure masking, that manifests when running Stencil parallel computations on an environment when failures are recovered locally. First, the delay propagation shape of one or multiple failures recovered locally is modeled to enable several analyses of the probability of different levels of failure masking under certain Stencil application behaviors. Our results indicate that failure masking is an extremely desirable effect at scale which manifestation is more evident and beneficial as the machine size or the failure rate increase.

A three-dimensional direct numerical simulation is conducted for a temporally evolving planar jet of n-heptane at a pressure of 40 atmospheres and in a coflow of air at 1100 K. At these conditions, n-heptane exhibits a two-stage ignition due to low- and high-temperature chemistry, which is reproduced by the global chemical model used in this study. The results show that ignition occurs in several overlapping stages and multiple modes of combustion are present. Low-temperature chemistry precedes the formation of multiple spatially localised high-temperature chemistry autoignition events, referred to as 'kernels'. These kernels form within the shear layer and core of the jet at compositions with short homogeneous ignition delay times and in locations experiencing low scalar dissipation rates. An analysis of the kernel histories shows that the ignition delay time is correlated with the mixing rate history and that the ignition kernels tend to form in vortically dominated regions of the domain, as corroborated by an analysis of the topology of the velocity gradient tensor. Once ignited, the kernels grow rapidly and establish edge flames where they envelop the stoichiometric isosurface. A combination of kernel formation (autoignition) and the growth of existing burning surface (via edge-flame propagation) contributes to the overall ignition process. An analysis of propagation speeds evaluated on the burning surface suggests that although the edge-flame speed is promoted by the autoignitive conditions due to an increase in the local laminar flame speed, edge-flame propagation of existing burning surfaces (triggered initially by isolated autoignition kernels) is the dominant ignition mode in the present configuration.

Abstract not provided.

The effect of differential molecular diffusion (DMD) in turbulent non-premixed flames is studied by examining two previously reported DNS of temporally evolving planar jet flames, one with CO/H2 as the fuel and the other with C2H4 as the fuel. The effect of DMD in the CO/H2 DNS flames in which H2 is part of fuel is found to behave similar to laminar flamelet, while in the C2H4 DNS flames in which H2 is not present in the fuel it is similar to laminar flamelet in early stages but becomes different from laminar flamelet later. The scaling of the effect of DMD with respect to the Reynolds number Re is investigated in the CO/H2 DNS flames, and an evident power law scaling (∼Re−a with a a positive constant) is observed. The scaling of the effect of DMD with respect to the Damköhler number Da is explored in both laminar counter-flow jet C2H4 diffusion flames and the C2H4 DNS flames. A power law scaling (∼Daa with a a positive constant) is clearly demonstrated for C2H4 nonpremixed flames.

Abstract not provided.

This article reports an analysis of the first detailed chemistry direct numerical simulation (DNS) of a high Karlovitz number laboratory premixed flame. The DNS results are first compared with those from laser-based diagnostics with good agreement. The subsequent analysis focuses on a detailed investigation of the flame area, its local thickness and their rates of change in isosurface following reference frames, quantities that are intimately connected. The net flame stretch is demonstrated to be a small residual of large competing terms: The positive tangential strain term and the negative curvature stretch term. The latter is found to be driven by flame speed-curvature correlations and dominated in net by low probability highly curved regions. Flame thickening is demonstrated to be substantial on average, while local regions of flame thinning are also observed. The rate of change of the flame thickness (as measured by the scalar gradient magnitude) is demonstrated, analogously to flame stretch, to be a competition between straining tending to increase gradients and flame speed variations in the normal direction tending to decrease them. The flame stretch and flame thickness analyses are connected by the observation that high positive tangential strain rate regions generally correspond with low curvature regions; these regions tend to be positively stretched in net and are relatively thinner compared with other regions. High curvature magnitude regions (both positive and negative) generally correspond with lower tangential strain; these regions are in net negatively stretched and thickened substantially.

Abstract not provided.

The flame structure corresponding to lean hydrogen–air premixed flames in intense sheared turbulence in the thin reaction zone regime is quantified from flame thickness and conditional scalar dissipation rate statistics, obtained from recent direct numerical simulation data of premixed temporally-evolving turbulent slot jet flames [1]. It is found that, on average, these sheared turbulent flames are thinner than their corresponding planar laminar flames. Extensive analysis is performed to identify the reason for this counter-intuitive thinning effect. The factors controlling the flame thickness are analyzed through two different routes i.e., the kinematic route, and the transport and chemical kinetics route. The kinematic route is examined by comparing the statistics of the normal strain rate due to fluid motion with the statistics of the normal strain rate due to varying flame displacement speed or self-propagation. It is found that while the fluid normal straining is positive and tends to separate iso-scalar surfaces, the dominating normal strain rate due to self-propagation is negative and tends to bring the iso-scalar surfaces closer resulting in overall thinning of the flame. The transport and chemical kinetics route is examined by studying the non-unity Lewis number effect on the premixed flames. The effects from the kinematic route are found to couple with the transport and chemical kinetics route. In addition, the intermittency of the conditional scalar dissipation rate is also examined. It is found to exhibit a unique non-monotonicity of the exponent of the stretched exponential function, conventionally used to describe probability density function tails of such variables. The non-monotonicity is attributed to the detailed chemical structure of hydrogen-air flames in which heat release occurs close to the unburnt reactants at near free-stream temperatures.

In the present work, a direct numerical simulation (DNS) of an experimental high Karlovitz number (Ka) CH4/air piloted premixed flame was analyzed to study the inner structure and the stabilization mechanism of the turbulent flame. A reduced chemical mechanism for premixed CH4/air combustion with NOx based on GRI-Mech3.0 was used, including 268 elementary reactions and 28 transported species. The evolution of the stretch factor, I0, indicates that the burning rate per unit flame surface area is considerably reduced in the near field and exhibits a minimum at x/D = 8. Downstream, the burning rate gradually increases. The stretch factor is different between different species, suggesting the quenching of some reactions but not others. Comparison between the turbulent flame and strained laminar flames indicates that certain aspects of the mean flame structure can be represented surprisingly well by flamelets if changes in boundary conditions are accounted for and the strain rate of the mean flow is employed; however, the thickening of the flame due to turbulence is not captured. The spatial development of displacement speeds is studied at higher Ka than previous DNS. In contrast to almost all previous studies, the mean displacement speed conditioned on the flame front is negative in the near field, and the dominant contribution to the displacement speed is normal diffusion with the reaction contribution being secondary. Further downstream, reaction overtakes normal diffusion, contributing to a positive displacement speed. The negative displacement speed in the near field implies that the flame front situates itself in the pilot region where the inner structure of the turbulent flame is affected significantly, and the flame stabilizes in balance with the inward flow. Notably, in the upstream region of the turbulent flame, the main reaction contributing to the production of OH, H+O2⇔O+OH (R35), is weak. Moreover, oxidation reactions, H2+OH⇔H+H2O (R79) and CO+OH⇔CO2+H (R94), are influenced by H2O and CO2 from the pilot and are completely quenched. Hence, the entire radical pool of OH, H and O is affected. However, the fuel consumption layer remains comparably active and generates heat, mainly via the reaction CH4+OH⇔CH3+H2O (R93).

A two-dimensional direct numerical simulation of the ignition at diesel engine-relevant conditions was performed at 40 atm and at 900 K ambient temperature using dimethyl ether (DME) as the fuel with a 30 species reduced chemical mechanism. At these conditions similar to diesel fuel DME exhibited two-stage ignition. The low-temperature chemistry analysis revealed a "spotty" first-stage autoignition in lean regions which transitions to a diffusively supported cool-flame and then propagates up the local mixture fraction gradient towards richer regions. The cool-flame speed was much faster than can be attributed to spatial gradients in first-stage ignition delay time in homogeneous reactors and it caused a shortening of the second-stage ignition delay times compared to a homogeneous reactor wherein the shortening became more pronounced at richer mixtures. Multiple high-temperature ignition kernels were noted over a range of rich mixtures that are much richer than the homogeneous most reactive mixture and most kernels formed much earlier than suggested by the homogeneous ignition delay time of the corresponding local mixture.

Transported probability density function (TPDF) methods are an attractive modeling approach for turbulent flames as chemical reactions appear in closed form. However, molecular micro-mixing needs to be modeled and this modeling is considered a primary challenge for TPDF methods. In the present study, a new algebraic mixing rate model for TPDF simulations of turbulent premixed flames is proposed, which is a key ingredient in commonly used molecular mixing models. The new model aims to properly account for the transition in reactive scalar mixing rate behavior from the limit of turbulence-dominated mixing to molecular mixing behavior in flamelets. An a priori assessment of the new model is performed using direct numerical simulation (DNS) data of a lean premixed hydrogen–air jet flame. The new model accurately captures the mixing timescale behavior in the DNS and is found to be a significant improvement over the commonly used constant mechanical-to-scalar mixing timescale ratio model. An a posteriori TPDF study is then performed using the same DNS data as a numerical test bed. The DNS provides the initial conditions and time-varying input quantities, including the mean velocity, turbulent diffusion coefficient, and modeled scalar mixing rate for the TPDF simulations, thus allowing an exclusive focus on the mixing model. The new mixing timescale model is compared with the constant mechanical-to-scalar mixing timescale ratio coupled with the Euclidean Minimum Spanning Tree (EMST) mixing model, as well as a laminar flamelet closure by Pope and Anand (1984). It is found that the laminar flamelet closure is unable to properly capture the mixing behavior in the thin reaction zones regime while the constant mechanical-to-scalar mixing timescale model under-predicts the flame speed. The EMST model coupled with the new mixing timescale model provides the best prediction of the flame structure and flame propagation among the models tested, as the dynamics of reactive scalar mixing across different flame regimes are appropriately accounted for.

The increasing complexity of both scientific simulations and high-performance computing system architectures are driving the need for adaptive workflows, in which the composition and execution of computational and data manipulation steps dynamically depend on the evolutionary state of the simulation itself. Consider, for example, the frequency of data storage. Critical phases of the simulation should be captured with high frequency and with high fidelity for postanalysis; however, we cannot afford to retain the same frequency for the full simulation due to the high cost of data movement. We can instead look for triggers, indicators that the simulation will be entering a critical phase, and adapt the workflow accordingly. In this paper, we present a methodology for detecting triggers and demonstrate its use in the context of direct numerical simulations of turbulent combustion using S3D. We show that chemical explosive mode analysis (CEMA) can be used to devise a noise-tolerant indicator for rapid increase in heat release. However, exhaustive computation of CEMA values dominates the total simulation, and thus is prohibitively expensive. To overcome this computational bottleneck, we propose a quantile sampling approach. Our sampling-based algorithm comes with provable error/confidence bounds, as a function of the number of samples. Most importantly, the number of samples is independent of the problem size, and thus our proposed sampling algorithm offers perfect scalability. Our experiments on homogeneous charge compression ignition and reactivity controlled compression ignition simulations show that the proposed method can detect rapid increases in heat release, and its computational overhead is negligible. Our results will be used to make dynamic workflow decisions regarding data storage and mesh resolution in future combustion simulations.

With the goal of providing a more detailed fundamental understanding of ignition processes in diesel engines, this study reports analysis of a direct numerical simulation (DNS) database. In the DNS, a pseudo turbulent mixing layer of dimethyl ether (DME) at 400 K and air at 900 K is simulated at a pressure of 40 atmospheres. At these conditions, DME exhibits a two-stage ignition and resides within the negative temperature coefficient (NTC) regime of ignition delay times, similar to diesel fuel. The analysis reveals a complex ignition process with several novel features. Autoignition occurs as a distributed, two-stage event. The high-temperature stage of ignition establishes edge flames that have a hybrid premixed/autoignition flame structure similar to that previously observed for lifted laminar flames at similar thermochemical conditions. A combustion mode analysis based on key radical species illustrates the multi-stage and multi-mode nature of the ignition process and highlights the substantial modelling challenge presented by diesel combustion.

Dissipation spectra of velocity and reactive scalars—temperature and fuel mass fraction—in turbulent premixed flames are studied using direct numerical simulation data of a temporally evolving lean hydrogen-air premixed planar jet (PTJ) flame and a statistically stationary planar lean methane-air (SP) flame. The equivalence ratio in both cases was 0.7, the pressure 1 atm while the unburned temperature was 700 K for the hydrogen-air PTJ case and 300 K for methane-air SP case, resulting in data sets with a density ratio of 3 and 5, respectively. The turbulent Reynolds numbers for the cases ranged from 200 to 428.4, the Damköhler number from 3.1 to 29.1, and the Karlovitz number from 0.1 to 4.5. The dissipation spectra collapse when normalized by the respective Favre-averaged dissipation rates. However, the normalized dissipation spectra in all the cases deviate noticeably from those predicted by classical scaling laws for constant-density turbulent flows and bear a clear influence of the chemical reactions on the dissipative range of the energy cascade.

Abstract not provided.

A three-dimensional direct numerical simulation (DNS) of a turbulent lifted dimethyl ether (DME) slot jet flame was performed at elevated pressure to study interactions between chemical reactions with low-temperature heat release (LTHR), negative temperature coefficient (NTC) reactions and shear generated turbulence in a jet in a heated coflow. By conditioning on mixture fraction, local reaction zones and local heat release rate, the turbulent flame is revealed to exhibit a "pentabrachial" structure that was observed for a laminar DME lifted flame [Krisman et al., (2015)]. The propagation characteristics of the stabilization and triple points are also investigated. Potential stabilization points, spatial locations characterized by preferred temperature and mixture fraction conditions, exhibit autoignition characteristics with large reaction rate and negligible molecular diffusion. The actual stabilization point which coincides with the most upstream samples from the pool of potential stabilization points fovr each spanwise location shows passive flame structure with large diffusion. The propagation speed along the stoichiometric surface near the triple point is compared with the asymptotic value obtained from theory [Ruetsch et al., (1995)]. At stoichiometric conditions, the asymptotic and averaged DNS values of flame displacement speed deviate by a factor of 1.7. However, accounting for the effect of low-temperature species on the local flame speed increase, these two values become comparable. This suggests that the two-stage ignition influences the triple point propagation speed through enhancement of the laminar flame speed in a configuration where abundant low-temperature products from the first stage, low-temperature ignition are transported to the lifted flame by the high-velocity jet.