Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “machine leaning”

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.

At least 55 records · Page 3

A study of laser induced ignition of methane–air mixtures inside a Rapid Compression Machine

We present herein a fundamental study of laser ignition of methane/air mixtures at temperatures and pressures representative of an internal combustion engine. An Nd:YAG laser operating at λ = 1064 nm was used to ignite methane/air mixtures at equivalence ratios of 0.4 ≤ Φ ≤ 1 in a Rapid Compression Machine (RCM). Experiments were conducted to study the lean limit, minimum spark energy (MSE), and minimum ignition energy (MIE). The results show that laser ignition exhibits a stochastic behavior which must be interpreted statistically. A 90% probability of occurrence was used to evaluate the MSE and MIE, which resulted in MSE 90 =2.3 mJ and MIE 90 =7.2 mJ at an equivalence ratio Φ = 0.4 at compressed pressure and temperature of P comp = 29 bar and T comp = 750 K, respectively. The lean limit was characterized based on the fraction of chemical energy converted into thermal energy, which was determined by calculating the apparent rate of heat release as derived from RCM high speed pressure data. A lean limit for 90% chemical energy conversion was found to correspond to an equivalence ratio of 0.47 (T comp = 782 K). Schlieren photography was employed as a diagnostics tool to visualize the flame initiation and propagation inside the RCM.

42 ENGINEERING↗

Unit cell manufacture by blown powder directed energy deposition

Lattice structures manufactured by additive manufacturing show unique promise to alter existing components to improve the thermo-mechanical properties of the base structure. Here, in the present study, the manufacture of a self-supporting lattice structure by blown powder directed energy deposition was explored. Converging struts were manufactured using multiple lean angles to facilitate the self-supporting structure and to avoid machine-part collisions. The struts showed near uniform straightness relative to a fitted line with a maximum deviation of 1mm at a lean angle of 10°. Micro–hardness and porosity evaluation results showed no significant variation along the struts.

Additive manufacturing↗

Rapid Response Small Machining NNR Project 703025

This project was an effort to develop a machining area for small sized parts that is capable of delivering product with a quick response time. This entailed focusing efforts on leaning out specific work cells that would result in overall improvement to the entire machining area. This effort involved securing the most efficient available technologies for these areas. In the end, this incorporated preparing the small machining area for transformation to a new facility.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

EdgeAI: Machine learning via direct attached accelerator for streaming data processing at high shot rate x-ray free-electron lasers

We present a case for low batch-size inference with the potential for adaptive training of a lean encoder model. We do so in the context of a paradigmatic example of machine learning as applied in data acquisition at high data velocity scientific user facilities such as the Linac Coherent Light Source-II x-ray Free-Electron Laser. We discuss how a low-latency inference model operating at the data acquisition edge can capitalize on the naturally stochastic nature of such sources. We simulate the method of attosecond angular streaking to produce representative results whereby simulated input data reproduce high-resolution ground truth probability distributions. By minimizing the mean-squared error between the decoded output of the latent representation and the ground truth distributions, we ensure that the encoding layers and resulting latent representation maintains full fidelity for any downstream task, be it classification or regression. We present throughput results for data-parallel inference of various batch sizes, some with throughput exceeding 100 k images per second. We also show in situ training below 10 s per epoch for the full encoder–decoder model as would be relevant for streaming and adaptive real-time data production at our nation’s scientific light sources.

97 MATHEMATICS AND COMPUTING↗

SU-F-T-163: Improve Proton Therapy Efficiency: Report of a Workshop

Purpose: The technology of proton therapy, especially the pencil beam scanning technique, is evolving very quickly. However, the efficiency of proton therapy seems to lag behind conventional photon therapy. The purpose of the abstract is to report on the findings of a workshop on improvement of QA, planning and treatment efficiency in proton therapy. Methods: A panel of physicists, clinicians, and vendor representatives from over 18 institutions in the United States and internationally were convened in Knoxville, Tennessee in November, 2015. The panel discussed several topics on how to improve proton therapy efficiency, including 1) lean principle and failure mode and effects analysis, 2) commissioning and machine QA, 3) treatment planning, optimization and evaluation, 4) patient positioning and IGRT, 5) vendor liaison and machine availability, and 6) staffing, education and training. Results: The relative time needed for machine QA, treatment planning & check in proton therapy was found to range from 1 to 2.5 times of that in photon therapy. Current status in proton QA, planning and treatment was assessed. Key areas for efficiency improvement, such as elimination of unnecessary QA items or steps and development of efficient software or hardware tools, were identified. A white paper to summarize our findings is being written. Conclusion: It is critical to improve efficiency by developing reliable proton beam lines, efficient software tools on treatment planning, optimization and evaluation, and dedicated proton QA device. Conscious efforts and collaborations from both industry leaders and proton therapy centers are needed to achieve this goal and further advance the technology of proton therapy.

60 APPLIED LIFE SCIENCES↗

Ignition, Flame Progation, and End-Gas Autoignition Studies of Natural Gas/ EGR Blends in a Rapid Compression Machine

The development of high efficiency, spark ignited natural gas engines is currently limited by engine knock at high power density (at high compression ratio and/or elevated boost pressures) and misfire (at lean conditions and/or high exhaust gas recirculation levels). The knock and misfire limits are further confounded by the wide variety in fuel reactivity observed in “pipeline quality” natural gas. In this study, a rapid compression machine (RCM) was used to characterize the effects of variation in natural gas fuel reactivity and exhaust gas recirculation (EGR) on homogeneous ignition delay, flame propagation rate and end-gas autoignition (EGAI) propensity for stoichiometric natural gas/oxidizer/EGR blends. Pipeline quality natural gas with variable reactivity (68 < Methane Number < 95) was simulated using mixtures of methane, ethane, and propane. EGR was simulated with mixtures of Ar, CO2, CO, and NO. EGR substitution rates were varied from 0 to 30 mass percent. Ignition delay period under homogeneous autoignition conditions was measured at compressed pressures of 24.0 to 33.1 bar and compressed temperatures of 667 to 980 K. Flame propagation rate and EGAI propensity was measured at spark-ignition pressures of 29.9 to 32.9 bar and temperatures of 721 to 782 K.

Mohr, Jeffrey↗

Ignition, Flame Propagation, and End-Gas Autoignition Studies of Natural Gas/EGR Blends in a Rapid Compression Machine

The development of high efficiency, spark ignited natural gas engines is currently limited by engine knock at high power density (at high compression ratio and/or elevated boost pressures) and misfire (at lean conditions and/or high exhaust gas recirculation levels). The knock and misfire limits are further confounded by the wide variety in fuel reactivity observed in “pipeline quality” natural gas. In this study, a rapid compression machine (RCM) was used to characterize the effects of variation in natural gas fuel reactivity and exhaust gas recirculation (EGR) on homogeneous ignition delay, flame propagation rate and end-gas autoignition (EGAI) propensity for stoichiometric natural gas/oxidizer/EGR blends. Pipeline quality natural gas with variable reactivity (68 < Methane Number < 95) was simulated using mixtures of methane, ethane, and propane. EGR was simulated with mixtures of Ar, CO2, CO, and NO. EGR substitution rates were varied from 0 to 30 mass percent. Ignition delay period under homogeneous autoignition conditions was measured at compressed pressures of 24.0 to 33.1 bar and compressed temperatures of 667 to 980 K. Flame propagation rate and EGAI propensity was measured at spark-ignition pressures of 29.9 to 32.9 bar and temperatures of 721 to 782 K.

Ignition Delay, Flame Propagation Rate, Rapid Comp↗

Autoignition of CRC diesel surrogates at low temperature combustion conditions: Rapid compression machine experiments and modeling

As federal programs require increasingly stringent engine emissions and fuel economy standards, these ambitions can only be met if next-generation combustion technology is developed focusing on high-efficiency and low-emissions engines. Recent research has indicated the need to operate engines at higher compression ratios and with low temperature combustion (LTC) to achieve the needed gains in engine efficiency and reductions in emissions. Because there is a lack of understanding of the chemistry of diesel fuel components and their mixtures at these LTC conditions, this limits the ability to develop predictive chemical kinetic models that can be used to optimize engine combustion. The current study aims to fill in gaps in fundamental combustion data on surrogate fuel mixtures relevant to diesel fuels. Specifically, four multicomponent diesel surrogates formulated by the Coordinating Research Council (CRC) to emulate an ultra-low-sulfur research-grade #2 certification diesel fuel (CFA), namely V0a (4 components), V0b (5 components), V1 (8 components), and V2 (9 components), have been investigated in a rapid compression machine (RCM) through determination of total and first-stage ignition delay times. Autoignition characteristics of lean to rich fuel/O 2 /N 2 mixtures, for the four CRC surrogates and CFA, have been measured using an RCM at LTC relevant pressures and temperatures, in the ranges of 10–20 bar and 650–1000 K, respectively. The equivalence ratios have been varied by independently changing the oxygen mole fraction and the fuel mole fraction in the test mixtures, thereby illustrating the individual effects of oxygen concentration and fuel loading on diesel autoignition. Autoignition results of these four CRC surrogates are compared among them and with those of CFA. Some degree of agreement in autoignition response between each CRC surrogate and CFA is observed, while discrepancies are also identified and discussed. In addition, a detailed chemical kinetic model for diesel surrogates has been developed and validated against these newly-acquired RCM data. We find that this model shows reasonable agreement with the overall ignition delay time results of the current RCM experiments. Chemical kinetic analyses of the developed model were further conducted to help identify the reactions controlling the autoignition processes and the consumption of fuel components in CRC surrogates.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A detailed experimental study of n-propylcyclohexane autoignition in lean conditions

The autoignition chemistry of lean n-propylcyclohexane/''air'' mixtures ({phi} = 0.3, 0.4, 0.5) was investigated in a rapid compression machine at compressed gas temperatures ranging from 620 to 930 K and pressures ranging from 0.45 to 1.34 MPa. Cool flame and ignition delay times were measured. Cool flame delay times were found to follow an Arrhenius behavior, and a correlation including pressure and equivalence ratio dependences was deduced. The present ignition delay data were compared with recent experimental results and simulations from the available thermokinetic models in the literature. Negative temperature coefficient zones were observed when plotting ignition delay times versus compressed gas temperature. The oxidation products were identified and quantified during the ignition delay period. Formation pathways for the C{sub 9} bicyclic ethers and conjugate alkenes are proposed. The experimental data provide an extensive database to test detailed thermokinetic oxidation models. (author)

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A conceptual design of rotating board technique for delivering total skin electron therapy

Purpose: This study presents a novel technique in which a uniform radiation dose to the whole body, soles, and scalp vertex can be achieved in one electron beam treatment fraction. Methods: The patient was treated at a machine with a home-made rotating board. The patients were treated in two groups in the prone and supine positions by leaning onto an inner rotational board in the prone and supine positions. Each group can further be separated into two subgroups using tilting and rotational positions for treatment. Results: One of the beams was directed 15.5 deg. upward and 15.5 deg. downward from the horizontal axis to provide a field size of as large as 200 cm in height and 140 cm in width. An incline angle of 31.5 deg. anteriorly (forward) or posteriorly (backward) of the outer frame at an angle rotated 60 deg. clockwise or counterclockwise to the inner frame was found to be most appropriate. The output for the rotating board total skin electron therapy (RB-TSET) was 0.046 cGy/MU at ISD of 350 cm. The beam characteristics of the RB-TSET depth dose curves were R{sub 50}=2.48 cm, d{sub max}=0.7 cm, E{sub 0}=5.78 MeV, and R{sub p}=3.4 cm. Conclusions: The RB-TSET technique presented in this study is able to deliver a uniform radiation dose to the patient's skin surface, the scalp vertex, and soles of the feet all at one time, eliminating the trouble of having to further irradiate these two regions separately when using the Stanford six field technique.

61 RADIATION PROTECTION AND DOSIMETRY↗

Factorization Machine Learning for Disaggregation of Transmission Load Profiles with High Penetration of Behind-the-Meter Solar

The ever-growing high penetration of ubiquitously distributed energy resources, especially behind-the-meter solar (BTM) generations, has significant impacts on nodal load (i.e., net injection) profiles and consequently caused imperative operational challenges to system operators such as regional transmission organizations (RTOs). Illustrated by real-world nodal data and examples at PJM Interconnection, this paper first discusses the application and necessity of effectively extracting daily nodal load profiles in a non-intrusive manner. More importantly, a novel bi-level architecture, including Factorization Machines (FM) learning procedure has been proposed to effectively disaggregate not only one node but every node in an RTO service territory. Specifically, FM leaning is adopted to capture the interconnections between related features to better utilize the correlation between buses in the same region and between a single bus and the zonal load. The proposed bi-level technique is numerically validated using real-world, minute-level, normalized, and anonymized nodal data at PJM service territory.

behind the meter solar, load disaggregation, load ↗

FutureTense

Protective vaccines and reliable diagnostics are essential tools for controlling viral diseases. However, the efficacy of these tools can be diminished by mutations in viral genomes. The delay between the emergence of new viral strains and the redesign of vaccines and diagnostics allows for continued viral transmission. Is it possible to address this challenge by computationally predicting viral genome sequence evolution? Can we “future-proof” vaccines and diagnostics by targeting both current and anticipated future sequence variants? While predicting viral evolution is still an unsolved, “grand challenge” problem in biology, the large, and rapidly growing, number of SARS-CoV-2 genome sequences provide an opportunity to quantify the ability of machine learning to predict viral genome sequence evolution. Towards this end, we have developed a simple computational model for predicting viral evolution at the level of individual nucleotides. The key metric for quantifying the per-base, prediction accuracy for viral evolution is the Mann-Whitney U statistic (or, equivalently, the area under the receiver operator curve). Since the Mann-Whitney U statistic is not a differentiable function, existing deep leaning packages (like Pytorch and Keras/TensorFlow) are not useful, as they require that the accuracy metric/objective function be analytically differentiable with respect to the model parameters. To overcome this challenge, we have implemented custom software, “FutureTense”, that can train a machine learning model by maximizing the non-differentiable Mann-Whitney U statistic. This software trains a machine learning model by exploring along the direction of the discrete gradient of the Mann-Whitney U statistic in the model parameter space. Parallel computing and genome sequence-specific optimizations are used to accelerate model training. The resulting machine learning model learns the observed high C->U mutation rates in the SARS-CoV-2 genome (which are potentially induced by host defenses) and provides prediction accuracies that are significantly better than one would expect from random chance. While predicting viral evolution is still quite far from a solved problem, the surprising performance of this simple model gives hope that the accuracy of predicting viral genome evolution can be further increased by more sophisticated approaches.

Gans, Jason↗

Computational Investigation of a CO 2 Conversion Strategy via Diels–Alder Reaction in a Carbon Capture Solvent

Molecular-level insights into reactive separations are crucial for the design of new conversion pathways of carbon dioxide (CO 2 ). This work explores a postulated pathway that directs CO 2 to undergo inverse-electron-demand Diels–Alder reactions to produce heterocycles using the CO 2 chemically fixed on water-lean solvent molecules. Density functional theory calculations are applied to evaluate the lowest unoccupied molecular orbital (LUMO) energies of three types of reactants (1,3-butadiene, 1,3-cyclohexadiene, and 1,2,4,5-tetrazine) with various functional substituents. These calculations also provide a data set (5.8k data) for developing a machine learning model to efficiently predict LUMO energies. A computational screening of LUMO energies for an additional 47k diene and tetrazine candidates is performed, and a list of candidates with lowered LUMO energies by electron-withdrawing substituents is provided. These candidates are further examined by their reaction energy barriers computed from the interatomic potential or density functional theory. Two major energy barriers are identified, one for the proton transfer within the water-lean solvent and the other for the CO 2 transfer from the solvent molecule to the reactant candidate (diene or tetrazine). The functional substituents have a more significant impact on the second barrier but a very slight one on the first barrier. This exploratory work demonstrates a new possibility for guiding experimental efforts toward the chemical conversion of fixated CO 2 to value-added compounds.

Chemical reactions↗

THE EFFECT OF FUEL REACTIVITY AND EXHAUST GAS RECIRCULATION ON KNOCK PROPENSITY OF NATURAL GAS

The development of high efficiency, spark ignited natural gas engines is currently limited by engine knock at high compression ratio/elevated boost pressures and misfire at lean conditions/high exhaust gas recirculation (EGR) levels. The knock and misfire limits are further confounded by the wide variety in fuel reactivity observed in “pipeline quality” natural gas. In this study, a rapid compression machine was used to characterize the effects of EGR and variation in natural gas fuel reactivity on the homogeneous ignition delay, flame propagation rate, and end- gas autoignition propensity for stoichiometric natural gas/oxidizer/EGR blends. A reduced chemical kinetic mechanism was also developed to accurately model the homogeneous ignition delays measured in the Colorado State University rapid compression machine (CSU RCM). Pipeline quality natural gas with a range of chemical reactivity (68 < Methane Number < 95) was simulated using mixtures of CH4, C2H6, and C3H8. Exhaust gas recirculation gases were simulated with mixtures of Ar, CO2, CO, and NO at substitution rates of 0 to 30 mass percent. Ignition delay period under homogeneous autoignition conditions was measured at compressed pressures of 30.2 to 34.0 bar and compressed temperatures of 667 to 980 K. End-gas autoignition fraction and flame propagation rate were measured by initiating a laser spark in the center of the combustion chamber, after compression, at pressures of 30.7 to 32.7 bar and temperatures of 751 to 795 K. The results indicate that both fuel reactivity and the presence of reactive species (NO and CO) in the exhaust gas recirculation have a strong impact on end-gas autoignition fraction. A chemical kinetic mechanism was developed to predict homogeneous ignition delays for pipeline quality natural gas in a pressure and temperature range of 1-100 bar and 500-1000 K respectively. This mechanism accurately predicted measured homogeneous ignition delay in the RCM with a total average relative error of 11.0%.

Mohr, Jeffrey↗

Improvements to the Powder Processing of near-Final Shape alnico Magnets

Alnico permanent magnets (PMs), a recent PM system of interest as an attractive rare earth-free PM alternative, have advantageous high operating temperature and magnetic saturation with the potential for utilization in electric machines, e.g. interior-PM motors found in electric vehicles, if current directional solidification methods can be replaced by a true mass production approach. Recently, two unique alnico compositions, termed Full-Co and Co-lean, with improved coercivity, were gas atomized, compression molded, and vacuum sintered (4h at 1240°C) to high densities of 97.8% and 99.3%, respectively. However, the Co-lean remained fine grained isotropic magnets and the Full-Co grains were not textured, lowering magnetic strength in spite of attempts to grow large textured grains by a stress-biased solid state grain alignment method to convert them to high energy anisotropic magnets. It was hypothesized that oxidation during de-binding in air left many prior particle boundary oxides within the sintered microstructure that hindered grain growth and texturing during the stress-biased texturing procedure and prevented the desired abnormal grain growth (AGG). Here we explored a vacuum de-binding step that was linked (in-place) to vacuum sintering and found that the Co-lean exhibited faster uniform grain growth that doubled the average grain size (40 μm to 80 μm). Linked vacuum de-binding and sintering of Full-Co produced some AGG after only 1 h of 1240°C sintering. A new direction for promoting AGG (and stress-biased texturing) in alnico is being explored that utilizes a fundamental analysis of systems with second phase particles that either inhibit or boost grain growth. This effort explores the influence of vacuum de-binding linked to a series of lower sintering temperatures at a fixed time (4h) to see if oxide particle size and volume fraction can be changed to promote AGG conditions in alnico. Surprising qualitative results indicate that AGG may be promoted for vacuum sintering at less than 1200°C.

Rinko, Emily↗

DLSIA: Deep Learning for Scientific Image Analysis

DLSIA (Deep Learning for Scientific Image Analysis) is a Python-based machine learning library that empowers scientists and researchers across diverse scientific domains with a range of customizable convolutional neural network (CNN) architectures for a wide variety of tasks in image analysis to be used in downstream data processing. DLSIA features easy-to-use architectures, such as autoencoders, tunable U-Nets and parameter-lean mixed-scale dense networks (MSDNets). Additionally, this article introduces sparse mixed-scale networks (SMSNets), generated using random graphs, sparse connections and dilated convolutions connecting different length scales. For verification, several DLSIA-instantiated networks and training scripts are employed in multiple applications, including inpainting for X-ray scattering data using U-Nets and MSDNets, segmenting 3D fibers in X-ray tomographic reconstructions of concrete using an ensemble of SMSNets, and leveraging autoencoder latent spaces for data compression and clustering. As experimental data continue to grow in scale and complexity, DLSIA provides accessible CNN construction and abstracts CNN complexities, allowing scientists to tailor their machine learning approaches, accelerate discoveries, foster interdisciplinary collaboration and advance research in scientific image analysis.

97 MATHEMATICS AND COMPUTING↗

An experimental study of the autoignition characteristics of conventional jet fuel/oxidizer mixtures: Jet-A and JP-8

Ignition delay times of Jet-A/oxidizer and JP-8/oxidizer mixtures are measured using a heated rapid compression machine at compressed charge pressures corresponding to 7, 15, and 30 bar, compressed temperatures ranging from 650 to 1100 K, and equivalence ratios varying from 0.42 to 2.26. When using air as the oxidant, two oxidizer-to-fuel mass ratios of 13 and 19 are investigated. To achieve higher compressed temperatures for fuel lean mixtures (equivalence ratio of {proportional_to}0.42), argon dilution is also used and the corresponding oxidizer-to-fuel mass ratio is 84.9. For the conditions studied, experimental results show two-stage ignition characteristics for both Jet-A and JP-8. Variations of both the first-stage and overall ignition delays with compressed temperature, compressed pressure, and equivalence ratio are reported and correlated. It is noted that the negative temperature coefficient phenomenon becomes more prominent at relatively lower pressures. Furthermore, the first-stage-ignition delay is found to be less sensitive to changes in equivalence ratio and primarily dependent on temperature. (author)

33 ADVANCED PROPULSION SYSTEMS↗

Integrated Mission Simulation for Long Term Space Flight

The problems of evaluating pilot performance, pilot training, and the integration of the man-machine system for proper functioning become increasingly difficult as the length of projected space flights increases. Further, actual in-flight experience becomes harder to obtain and the major means for training and evaluation becomes simulation. The first major source of difficulty in any simulation stems from a basic requirement of leaning theory: that for increased proficiency of performance to occur within any biologically capable system, the stimuli which are the bases for the expected behavior must be presented in an amenable form to the learner. Further, the sequence of presentation of the pertinent stimuli must be such that the learner can form the appropriate concepts and interactions required for the ultimate use of the learned material. It can then be seen that two basic factors of learning complex skills are: 1) Fidelity of simulation of pertinent stimuli. 2) Appropriate sequencing of stimuli during learning for proper concept formation.

SPACE FLIGHT↗