The chemical kinetics ODEs arising from operator-split reactive-flow simulations were solved on GPUs using explicit integration algorithms. Nonstiff chemical kinetics of a hydrogen oxidation ...mechanism (9 species and 38 irreversible reactions) were computed using the explicit fifth-order Runge–Kutta–Cash–Karp method, and the GPU-accelerated version performed faster than single- and six-core CPU versions by factors of 126 and 25, respectively, for 524,288 ODEs. Moderately stiff kinetics, represented with mechanisms for hydrogen/carbon-monoxide (13 species and 54 irreversible reactions) and methane (53 species and 634 irreversible reactions) oxidation, were computed using the stabilized explicit second-order Runge–Kutta–Chebyshev (RKC) algorithm. The GPU-based RKC implementation demonstrated an increase in performance of nearly 59 and 10 times, for problem sizes consisting of 262,144 ODEs and larger, than the single- and six-core CPU-based RKC algorithms using the hydrogen/carbon-monoxide mechanism. With the methane mechanism, RKC-GPU performed more than 65 and 11 times faster, for problem sizes consisting of 131,072 ODEs and larger, than the single- and six-core RKC-CPU versions, and up to 57 times faster than the six-core CPU-based implicit VODE algorithm on 65,536 ODEs. In the presence of more severe stiffness, such as ethylene oxidation (111 species and 1566 irreversible reactions), RKC-GPU performed more than 17 times faster than RKC-CPU on six cores for 32,768 ODEs and larger, and at best 4.5 times faster than VODE on six CPU cores for 65,536 ODEs. With a larger time step size, RKC-GPU performed at best 2.5 times slower than six-core VODE for 8192 ODEs and larger. Therefore, the need for developing new strategies for integrating stiff chemistry on GPUs was discussed.
A novel implementation for the skeletal reduction of large detailed reaction mechanisms using the directed relation graph with error propagation and sensitivity analysis (DRGEPSA) is developed and ...presented with examples for three hydrocarbon components,
n-heptane,
iso-octane, and
n-decane, relevant to surrogate fuel development. DRGEPSA integrates two previously developed methods, directed relation graph-aided sensitivity analysis (DRGASA) and directed relation graph with error propagation (DRGEP), by first applying DRGEP to efficiently remove many unimportant species prior to sensitivity analysis to further remove unimportant species, producing an optimally small skeletal mechanism for a given error limit. It is illustrated that the combination of the DRGEP and DRGASA methods allows the DRGEPSA approach to overcome the weaknesses of each, specifically that DRGEP cannot identify all unimportant species and that DRGASA shields unimportant species from removal. Skeletal mechanisms for
n-heptane and
iso-octane generated using the DRGEP, DRGASA, and DRGEPSA methods are presented and compared to illustrate the improvement of DRGEPSA. From a detailed reaction mechanism for
n-alkanes covering
n-octane to
n-hexadecane with 2115 species and 8157 reactions, two skeletal mechanisms for
n-decane generated using DRGEPSA, one covering a comprehensive range of temperature, pressure, and equivalence ratio conditions for autoignition and the other limited to high temperatures, are presented and validated. The comprehensive skeletal mechanism consists of 202 species and 846 reactions and the high-temperature skeletal mechanism consists of 51 species and 256 reactions. Both mechanisms are further demonstrated to well reproduce the results of the detailed mechanism in perfectly-stirred reactor and laminar flame simulations over a wide range of conditions. The comprehensive and high-temperature
n-decane skeletal mechanisms are included as supplementary material with this article.
Smoldering combustion plays an important role in forest and wildland fires. Fires from smoldering combustion can last for long periods of time, emit more pollutants, and be difficult to extinguish. ...This makes the study of smoldering in woody fuels and forest duff important. Cellulose, hemicellulose, and lignin are the major constituents in these type of fuels, in different proportions for different fuels. In this article, we developed a 1-D model using the open-source software Gpyro to study the smoldering combustion of cellulose and hemicellulose mixtures. We first validated our simulations against experimentally obtained values of propagation speed for mixtures with fuel compositions including 100%, 75%, 50%, and 25% cellulose, with the remaining proportion of hemicellulose. Then, we studied the effects of varying fuel composition, density, and moisture content on smoldering combustion. We found that propagation speed of smoldering increases with decreasing density and increasing hemicellulose content, which we attribute to the role of oxygen diffusion. Propagation speed increases with moisture content for pure cellulose up to a certain limiting value, after which the propagation speed drops by up to 70%. The mean peak temperature of smoldering increases with hemicellulose content and density, and decreases with increasing moisture content.
Liquid transportation fuels require costly and time-consuming tests to characterize metrics, such as Research Octane Number (RON) for gasoline. If fuel sale restrictions requiring use of standard ...Cooperative Fuel Research (CFR) testing procedures do not apply, these tests may be avoided by using multivariate statistical models to predict RON and other quantities. Existing techniques inform these models using information about existing, similar fuels—for example, training a model for gasoline RON with a large number of characterized gasoline samples. While this yields the most accurate predictive models for these fuels, this approach lacks the ability to predict characteristics of fuels outside the training data set. Here we show that an accurate statistical model for the RON of gasoline and gasoline-like fuels can be constructed by ensuring the representation of key functional groups in the spectroscopic data set are used to train the model. We found that a principal component regression model for RON based on IR absorbance and informed using neat and 134 mixtures of n-heptane, isooctane, toluene, ethanol, methylcyclohexane, and 1-hexene could predict RON for the 10 Coordinating Research Council (CRC) Fuels for Advanced Combustion Engine (FACE) gasolines and 12 FACE gasoline blends with ethanol within 34.8±36.1 on average and 51.2 in the worst case. We next studied the effect of adding 28 additional minor components found in the FACE gasolines to the statistical model, and determined that it was necessary to add additional representatives of the branched alkane and aromatics classes to reduce model error. For example, adding 2,3-dimethylpentane and xylene to the previous model allowed it to predict RON for the 22 target fuels within 0.3±4.4 on average and 7.9 in the worst case. However, we determined that the specific choice of fuel in those classes mattered less than ensuring the representation of the relevant functional group. This work builds upon previous efforts by creating models informed by neat and surrogate fuels—rather than complex real fuels—that could predict the performance of complex unknown fuels.
Strategies and recommendations for performing skeletal reductions of multicomponent surrogate fuels are presented, through the generation and validation of skeletal mechanisms for a three-component ...toluene reference fuel. Using the directed relation graph with error propagation and sensitivity analysis method followed by a further unimportant reaction elimination stage, skeletal mechanisms valid over comprehensive and high-temperature ranges of conditions were developed at varying levels of detail. These skeletal mechanisms were generated based on autoignition simulations, and validation using ignition delay predictions showed good agreement with the detailed mechanism in the target range of conditions. When validated using phenomena other than autoignition, such as perfectly stirred reactor and laminar flame propagation, tight error control or more restrictions on the reduction during the sensitivity analysis stage were needed to ensure good agreement. In addition, tight error limits were needed for close prediction of ignition delay when varying the mixture composition away from that used for the reduction. In homogeneous compression-ignition engine simulations, the skeletal mechanisms closely matched the point of ignition and accurately predicted species profiles for lean to stoichiometric conditions. Furthermore, the efficacy of generating a multicomponent skeletal mechanism was compared to combining skeletal mechanisms produced separately for neat fuel components; using the same error limits, the latter resulted in a larger skeletal mechanism size that also lacked important cross reactions between fuel components. Based on the present results, general guidelines for reducing detailed mechanisms for multicomponent fuels are discussed.
Machine learning has proven to be a powerful tool for accelerating biofuel development. Although numerous models are available to predict a range of properties using chemical descriptors, there is a ...trade-off between interpretability and performance. Neural networks provide predictive models with high accuracy at the expense of some interpretability, while simpler models such as linear regression often lack in accuracy. In addition to model architecture, feature selection is also critical for developing interpretable and accurate predictive models. We present a method for systematically selecting molecular descriptor features and developing interpretable machine learning models without sacrificing accuracy. Our method simplifies the process of selecting features by reducing feature multicollinearity and enables discoveries of new relationships between global properties and molecular descriptors. To demonstrate our approach, we developed models for predicting melting point, boiling point, flash point, yield sooting index, and net heat of combustion with the help of the Tree-based Pipeline Optimization Tool (TPOT). For training, we used publicly available experimental data for up to 8351 molecules. Our models accurately predict various molecular properties for organic molecules (mean absolute percent error (MAPE) ranges from 3.3% to 10.5%) and provide a set of features that are well-correlated to the property. This method enables researchers to explore sets of features that significantly contribute to the prediction of the property, offering new scientific insights. To help accelerate early stage biofuel research and development, we also integrated the data and models into a open-source, interactive web tool.
•Developed method for selecting chemical descriptors and minimizing collinearity•Trained five property prediction models using diverse data sets•Models are interpretable and yield excellent peformance•Feature importances are consistent and agree with previous research•Webtool available at feedstock-to-function.lbl.gov
Implementing multicomponent diffusion models in numerical combustion studies is computationally expensive; to reduce cost, numerical simulations commonly use mixture-averaged diffusion treatments or ...simpler models. However, the accuracy and appropriateness of mixture-averaged diffusion has not been verified for three-dimensional, turbulent, premixed flames. In this study we evaluated the role of multicomponent mass diffusion in premixed, three-dimensional high Karlovitz-number hydrogen, n-heptane, and toluene flames, representing a range of fuel Lewis numbers. We also studied a premixed, unstable two-dimensional hydrogen flame due to the importance of diffusion effects in such cases. Our comparison of diffusion flux vectors revealed differences of 10–20% on average between the mixture-averaged and multicomponent diffusion models, and greater than 40% in regions of high flame curvature. Overall, however, the mixture-averaged model produces small differences in diffusion flux compared with global turbulent flame statistics. To evaluate the impact of these differences between the two models, we compared normalized turbulent flame speeds and conditional means of species mass fraction and source term. We found differences of 5–20% in the mean normalized turbulent flame speeds, which seem to correspond to differences of 5–10% in the peak fuel source terms. Our results motivate further study into whether the mixture-averaged diffusion model is always appropriate for DNS of premixed turbulent flames.
The importance of graph search algorithm choice to the directed relation graph with error propagation (DRGEP) method is studied by comparing basic and modified depth-first search, basic and
...R-value-based breadth-first search (RBFS), and Dijkstra’s algorithm. By using each algorithm with DRGEP to produce skeletal mechanisms from a detailed mechanism for
n-heptane with randomly-shuffled species order, it is demonstrated that only Dijkstra’s algorithm and RBFS produce results independent of species order. In addition, each algorithm is used with DRGEP to generate skeletal mechanisms for
n-heptane covering a comprehensive range of autoignition conditions for pressure, temperature, and equivalence ratio. Dijkstra’s algorithm combined with a coefficient scaling approach is demonstrated to produce the most compact skeletal mechanism with a similar performance compared to larger skeletal mechanisms resulting from the other algorithms. The computational efficiency of each algorithm is also compared by applying the DRGEP method with each search algorithm on the large detailed mechanism for
n-alkanes covering
n-octane to
n-hexadecane with 2115 species and 8157 reactions. Dijkstra’s algorithm implemented with a binary heap priority queue is demonstrated as the most efficient method, with a CPU cost two orders of magnitude less than the other search algorithms.
Peer review of research articles is a core part of our scholarly communication system. In spite of its importance, the status and purpose of peer review is often contested. What is its role in our ...modern digital research and communications infrastructure? Does it perform to the high standards with which it is generally regarded? Studies of peer review have shown that it is prone to bias and abuse in numerous dimensions, frequently unreliable, and can fail to detect even fraudulent research. With the advent of web technologies, we are now witnessing a phase of innovation and experimentation in our approaches to peer review. These developments prompted us to examine emerging models of peer review from a range of disciplines and venues, and to ask how they might address some of the issues with our current systems of peer review. We examine the functionality of a range of social Web platforms, and compare these with the traits underlying a viable peer review system: quality control, quantified performance metrics as engagement incentives, and certification and reputation. Ideally, any new systems will demonstrate that they out-perform and reduce the biases of existing models as much as possible. We conclude that there is considerable scope for new peer review initiatives to be developed, each with their own potential issues and advantages. We also propose a novel hybrid platform model that could, at least partially, resolve many of the socio-technical issues associated with peer review, and potentially disrupt the entire scholarly communication system. Success for any such development relies on reaching a critical threshold of research community engagement with both the process and the platform, and therefore cannot be achieved without a significant change of incentives in research environments.
We present version 4.0 of the atmospheric chemistry box model CAABA/MECCA that now includes a number of new features: (i) skeletal mechanism reduction, (ii) the Mainz Organic Mechanism (MOM) chemical ...mechanism for volatile organic compounds, (iii) an option to include reactions from the Master Chemical Mechanism (MCM) and other chemical mechanisms, (iv) updated isotope tagging, and (v) improved and new photolysis modules (JVAL, RADJIMT, DISSOC). Further, when MECCA is connected to a global model, the new feature of coexisting multiple chemistry mechanisms (PolyMECCA/CHEMGLUE) can be used. Additional changes have been implemented to make the code more user-friendly and to facilitate the analysis of the model results. Like earlier versions, CAABA/MECCA-4.0 is a community model published under the GNU General Public License.