SEARCH · Engineering Papers

Results for “distributed inference”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

All sources NASA NTRS DOE OSTI

NASA NTRS1974

Exploratory tests of two strut fuel injectors for supersonic combustion

Results of supersonic mixing and combustion tests performed with two simple strut injector configurations, one with parallel injectors and one with perpendicular injectors, are presented and analyzed. Good agreement is obtained between static pressure measured on the duct wall downstream of the strut injectors and distributions obtained from one-dimensional calculations. Measured duct heat load agrees with results of the one-dimensional calculations for moderate amounts of reaction, but is underestimated when large separated regions occur near the injection location. For the parallel injection strut, good agreement is obtained between the shape of the injected fuel distribution inferred from gas sample measurements at the duct exit and the distribution calculated with a multiple-jet mixing theory. The overall fraction of injected fuel reacted in the multiple-jet calculation closely matches the amount of fuel reaction necessary to match static pressure with the one-dimensional calculation. Gas sample measurements with the perpendicular injection strut also give results consistent with the amount of fuel reaction in the one-dimensional calculation.

Anderson, G. Y.↗

NASA NTRS1981

Information theory lateral density distribution for Earth inferred from global gravity field

Information Theory Inference, better known as the Maximum Entropy Method, was used to infer the lateral density distribution inside the Earth. The approach assumed that the Earth consists of indistinguishable Maxwell-Boltzmann particles populating infinitesimal volume elements, and followed the standard methods of statistical mechanics (maximizing the entropy function). The GEM 10B spherical harmonic gravity field coefficients, complete to degree and order 36, were used as constraints on the lateral density distribution. The spherically symmetric part of the density distribution was assumed to be known. The lateral density variation was assumed to be small compared to the spherically symmetric part. The resulting information theory density distribution for the cases of no crust removed, 30 km of compensated crust removed, and 30 km of uncompensated crust removed all gave broad density anomalies extending deep into the mantle, but with the density contrasts being the greatest towards the surface (typically + or 0.004 g cm 3 in the first two cases and + or - 0.04 g cm 3 in the third). None of the density distributions resemble classical organized convection cells. The information theory approach may have use in choosing Standard Earth Models, but, the inclusion of seismic data into the approach appears difficult.

Rubincam, D. P.↗

DOE OSTI2026

IntraShuffler: A Privacy Preserving Framework for Heterogeneous DP Federated Learning

Heterogeneous Differential Privacy (HDP) in Federated Learning (FL) allows clients to select individual privacy budgets () according to institutional policies and data sensitivity. In practice, many HDP-FL systems employ -aware server aggregation to improve model utility by re-weighting client updates according to their declared privacy budgets. However, gradient updates in FL retain structural patterns induced by non-independent and identically-distributed (non-IID) data, and these additional signals exposed by -aware aggregation create new opportunities for inference by an honest-but-curious server. In this work, we first show that a server equipped with gradient denoising and surrogate modeling can mount a Privacy Inference Attack that infers distributional attributes of clients and links updates from the same client across training rounds, measured via surrogate inference accuracy and linkage success, under realistic knowledge constraints. The Shuffle-Model has been widely studied as a defense against such inference risks by anonymizing update sources, but it is fundamentally incompatible with HDP-FL -aware aggregation. To address this challenge, we propose IntraShuffler, a middleware defense framework designed for HDP-FL systems. IntraShuffler introduces a privacy-aware shuffling mechanism that groups clients into privacy-compatible buckets and performs parameter-level shuffling within each bucket to disrupt persistent gradient structure while preserving -aware aggregation. Experiments across four different datasets show that IntraShuffler reduces gradient recoverability by over 60% and decreases surrogate inference accuracy from 0.78 to 0.33 while maintaining comparable model utility across multiple FL aggregation rules.

Riya, Farhin Farhad [ORNL]↗

DOE OSTI2022

Inference-Optimized AI and High Performance Computing for Gravitational Wave Detection at Scale

We introduce an ensemble of artificial intelligence models for gravitational wave detection that we trained in the Summit supercomputer using 32 nodes, equivalent to 192 NVIDIA V100 GPUs, within 2 h. Once fully trained, we optimized these models for accelerated inference using NVIDIA TensorRT. We deployed our inference-optimized AI ensemble in the ThetaGPU supercomputer at Argonne Leadership Computer Facility to conduct distributed inference. Using the entire ThetaGPU supercomputer, consisting of 20 nodes each of which has 8 NVIDIA A100 Tensor Core GPUs and 2 AMD Rome CPUs, our NVIDIA TensorRT-optimized AI ensemble processed an entire month of advanced LIGO data (including Hanford and Livingston data streams) within 50 s. Our inference-optimized AI ensemble retains the same sensitivity of traditional AI models, namely, it identifies all known binary black hole mergers previously identified in this advanced LIGO dataset and reports no misclassifications, while also providing a 3X inference speedup compared to traditional artificial intelligence models. We used time slides to quantify the performance of our AI ensemble to process up to 5 years worth of advanced LIGO data. In this synthetically enhanced dataset, our AI ensemble reports an average of one misclassification for every month of searched advanced LIGO data. We also present the receiver operating characteristic curve of our AI ensemble using this 5 year long advanced LIGO dataset. This approach provides the required tools to conduct accelerated, AI-driven gravitational wave detection at scale.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1975

Molecular hydrogen in the galaxy and galactic gamma rays

Estimates made of column densities of H2 at galactic longitude equals 0 deg indicate that H2 is far more abundant than HI in the inner galaxy and is the key to an explanation of the gamma-ray observations. This is also reflected in the correlation of galactic longitude and latitude distributions of gamma-rays and molecular clouds. Particularly strong evidence is found from the galactic survey of CO emission at 2.64 mm. The cosmic-ray distribution inferred from the calculations is not uniform but only weakly dependent on the total gas distribution in the inner galaxy.

Stecker, F. W.↗

NASA NTRS1973

Neutral and ion-exospheres in the solar wind with applications to Mercury

Construction of a model neutral and ion exosphere for a planet weakly interacting with the solar wind. The model is constructed in general terms and is then specialized to possible neutral and ion exospheres for the planet Mercury. The neutral exosphere model allows for density and temperature variations and for rotation at the exobase. The ion exosphere is produced by ionization of the neutral exosphere in the solar wind, and its density distribution is obtained by solving the continuity equation in the drift approximation. Applying to Mercury a surface temperature distribution inferred from infrared data and a vanishing bound neutral flux at the base, He and He(+) density distributions are found. When the He atmosphere of Mercury is due entirely to the surface bombardment by solar wind He(2+), the resulting He(+) density is found to vary from 0.15 to 0.001 per cu cm over the range from 1.5 to 5 planetocentric radii on the dayside. These densities are found to be detectable by typical solar-wind plasma instruments.

Hartle, R. E.↗

NASA NTRS1989

The shape of the mass distribution in M31 from its globular cluster system

The velocity dispersion and rotation velocity of the M31 globular cluster system depend on the relative division of mass between the flat disk and a spherically symmetric halo. Using the tensor virial theorem, it is shown in detail how the mass ratio can be constrained. Radial velocities have been collected for 149 globular clusters in M31. With no assumptions about the isotropy of the velocity distribution, the globular cluster kinematics are consistent with the mass distribution inferred by the rotation curve but otherwise place no strong constraints on the relative division of the mass. If the velocity distribution is isotropic, models with the disk mass ranging between 1/2 and 1 times its maximum possible value are marginally favored.

Kent, Stephen M.↗

NASA NTRS1973

Mechanisms for Mars dust storms.

Characteristics of the Mars global dust storm are reviewed. At the Mariner 9 encounter, the dust consisted of highly absorbing particles distributed rather uniformly up to great height (about 50 km). These observations together with temperature distributions inferred from the Mariner 9 IRIS by Hanel et al. (1972) are used to estimate global wind systems during the dust storm. The global distribution and direction of light surface streaks indicate that the axisymmetric circulation was a dominant part of flow during the dust storm. The axisymmetric winds may become strong enough to raise dust over wide areas of Mars' tropics under unusual conditions: the incoming solar radiation must be near its seasonal maximum, the static stability must be low, and the atmosphere must be able to absorb and re-emit a sizeable fraction of the incoming radiation. Strong winds around the periphery of the retreating south polar cap would be driven by the temperature gradient at the cap edge and by the mass outflow due to subliming CO2. These polar winds could generate local dust storms, raising the general level of dustiness, and providing the conditions necessary for onset of a global dust storm.

Leovy, C. B.↗

NASA NTRS1986

Transient medium-scale wave activity in the summer stratosphere

Observations of medium-scale transient wave activity in the middle-latitude Southern Hemisphere summer stratosphere are interpreted using isentropic potential vorticity (IPV) distributions inferred from temperature data from the Nimbus-7 Limb Infrared Monitor of the Stratosphere (LIMS) instrument. Despite problems with errors in reference-level synoptic height information, the derived IPV distributions clearly exhibit dynamically coherent signatures of eddy mixing due to medium-scale disturbances in the lower-middle stratosphere during early January 1979. The IPV distributions are shown to be strongly correlated with those of a contemporaneous set of isentropic maps of quasi-conservative tracers such as ozone and nitric acid. The results suggest that IPV maps inferred from satellite temperature data provide a useful diagnostic for studies of large-scale dynamics and/or transport in the extratropical stratosphere. Furthermore, the comparatively high vertical resolution of the LIMS data has enabled the vertical attenuation of Southern Hemisphere wave activity to be diagnosed in greater detail than could be achieved from nadir-viewing satellite information.

Miles, T.↗

DOE OSTI2022

Soot-particle core-shell and fractal structures from small-angle X-ray scattering measurements in a flame

We have characterized soot particles measured in situ in a laminar co-flow ethylene-air diffusion flame using small-angle X-ray scattering (SAXS). The analysis includes temperature measurements made with coherent anti-Stokes Raman spectroscopy (CARS) and complements soot volume-fraction and maturity measurements made with laser-induced incandescence (LII). We compared the results of fits to the SAXS measurements using a unified model and a fractal core-shell model. Power-law parameters yielded by the unified model indicate that aggregates of primary particles are in the mass-fractal regime, whereas the primary particles are in the surface-fractal regime in the middle of the flame. Higher and lower in the flame, the primary-particle power-law parameter approaches 4, suggesting smooth primary particles. Furthermore, these trends are consistent with fits using the fractal core-shell model, which indicate that particles have an established core-shell structure in the middle of the flame and are internally homogeneous at higher and lower heights in the flame. Primary-particle size distributions derived using the fractal core-shell model demonstrate excellent agreement with distributions inferred from transmission electron microscopy (TEM) images in the middle of the flame. Higher in the flame, a second small mode appears in the size distributions, suggesting particle fragmentation during oxidation. Surface oxidation would explain (1) aggregate fragmentation and (2) loss of core-shell structure leading to smoother primary-particle surfaces by removal of carbon overlayers. SAXS measurements are much more sensitive to incipient and young soot particles than LII and demonstrate significant volume fraction from particles low in the flame where the LII signal is negligible.

42 ENGINEERING↗

DOE OSTI2022

Density reconstruction in convergent high-energy-density systems using x-ray radiography and Bayesian inference

X-ray radiography is a technique frequently used to diagnose convergent high-energy-density (HED) systems, such as inertial confinement fusion implosion, and provides unique information that is not available through self-emission measurements. In this work, we investigate the scope and limits of that information using a radiography simulation combined with the Bayesian inference workflow. The accuracy of density reconstruction from simulated radiographs of spherical implosions driven with 27 kJ laser energy is assessed, including the increase or decrease in accuracy due to the addition of Lagrangian marker layers, Poisson noise, and improved prior information. This work is the first to present the full uncertainty distributions inferred from radiography analysis in HED systems and demonstrates the importance of constructing the full posterior probability density, as opposed to a point estimate, due to the modal structure of the likelihood surface introduced by typical experimental noise sources. This general methodology can be used both for robust analysis of radiographic data and for an improved design of radiography experiments by modeling the full experimental system.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

NASA NTRS1986

Optical probing of comet Halley from the Giotto spacecraft

The Halley optical probe experiment (HOPE) aboard the Giotto spacecraft has provided the first measurements of the optical properties of the dust and of some gaseous species (CN, C2, CO/+/ and OH) from inside the coma of comet Halley. The dust spatial distribution, inferred from sunlight scattering, obeys an r exp -2 law for distances from the nucleus r greater than about 2,000 km. The more rapid increase of dust density with distance observed in the innermost coma probably reflects the presence of a jet, or a surge of activity within a few hours before closest approach. The CN and OH signals increase more slowly than the dust signals, but faster than predicted by a simple model (of solar-excited resonance fluorescence) for the distribution of these species. This effect is probably caused in part by the dust contribution, as well as by the production from parent molecules of OH and CN in an excited state.

Levasseur-Regourd, A. C.↗

DOE OSTI2020

Strong Lensing Model of SPT-CL J0356–5337, a Major Merger Candidate at Redshift 1.0359

We present an analysis of the mass distribution inferred from strong lensing by SPT-CL J0356-5337, a cluster of galaxies at redshift $z=1.0359$ revealed in the follow-up of the SPT-SZ clusters. The cluster has an Einstein radius of ${\theta }_{{\rm{E}}}\simeq $ 14'' for a source at z = 3 and a mass within 500 kpc of ${M}_{500\mathrm{kpc}}=4.0\pm 0.8\times {10}^{14}\,$ ${M}_{\odot }$ . Our spectroscopic identification of three multiply imaged systems ($z=2.363$, $z=2.364$, and $z=3.048$), combined with HSTF606W-band imaging allows us to build a strong lensing model for this cluster with an rms of $\leqslant 0\buildrel{\prime\prime}\over{.} 3$. Our modeling reveals a two-component mass distribution in the cluster. One mass component is dominated by the brightest cluster Galaxy and the other component, separated by ~170 kpc, contains a group of eight red elliptical galaxies confined in a ~9'' (~70 kpc) diameter circle. We estimate the mass ratio between the two components to be between 1:1.25 and 1:1.58. In addition, spectroscopic data reveal that these two near-equal mass cores have only a small velocity difference of ~300 km s-1 between the two components. This small radial velocity difference suggests that most of the relative velocity takes place in the plane of the sky, and implies that SPT-CL J0356-5337 is a major merger with a small impact parameter seen face-on. We also assess the relative contributions of Galaxy-scale halos to the overall mass of the core of the cluster and find that within 800 kpc from the brightest cluster Galaxy about 27% of the total mass can be attributed to visible and dark matter associated with galaxies, whereas only 73% of the total mass in the core comes from cluster-scale dark matter halos.

79 ASTRONOMY AND ASTROPHYSICS↗

NASA NTRS1996

Lithologic Distribution and Geologic History of the Apollo 17 Site: The Record in Soils and Small Rock Particles from the Highland Massifs

Through analysis by instrumental neutron activation (INAA) of 789 individual lithic fragments from the 2 mm-4 mm grain-size fractions of five Apollo 17 soil samples (72443, 72503, 73243, 76283, and 76503) and petrographic examination of a subset, we have determined the diversity and proportions of rock types recorded within soils from the highland massifs. The distribution of rock types at the site, as recorded by lithic fragments in the soils, is an alternative to the distribution inferred from the limited number of large rock samples. The compositions and proportions of 2 mm-4 mm fragments provide a bridge between compositions of less than 1 mm fines and types and proportions of rocks observed in large collected breccias and their clasts. The 2 mm-4 mm fraction of soil from South Massif, represented by an unbiased set of lithic fragments from station-2 samples 72443 and 72503, consists of 71% noritic impact-melt breccia, 7% Incompatible-Trace-Element-(ITE)-poor highland rock types (mainly granulitic breccias), 19% agglutinates and regolith breccias, 1% high-Ti mare basalt, and 2% others (very-low-Ti (VLT) basalt, monzogabbro breccia, and metal). In contrast, the 2 mm - 4 mm fraction of a soil from the North Massif, represented by an unbiased set of lithic fragments from station-6 sample 76503, has a greater proportion of ITE-poor highland rock types and mare-basalt fragments: it consists of 29% ITE-poor highland rock types (mainly granulitic breccias and troctolitic anorthosite), 25% impact-melt breccia, 13% high-Ti mare basalt, 31 % agglutinates and regolith breccias, 1% orange glass and related breccia, and 1% others. Based on a comparison of mass- weighted mean compositions of the lithic fragments with compositions of soil fines from all Apollo 17 highland stations, differences between the station-2 and station-6 samples are representative of differences between available samples from the two massifs. From the distribution of different rock types and their compositions, we conclude the following: (1) North-Massif and South-Massif soil samples differ significantly in types and proportions of ITE-poor highland components and ITE-rich impact-melt-breccia components. These differences reflect crudely layered massifs and known local geology. The greater percentage of impact-melt breccia in the South- Massif light-mantle soil stems from derivation of the light mantle from the top of the massif, which apparently is richer in noritic impact-melt breccia than are lower parts of the massifs. (2) At station 2, the 2 mm-4 mm grain-size fraction is enriched in impact-melt breccias compared to the less than 1 mm fraction, suggesting that the <1 mm fraction within the light mantle has a greater proportion of lithologies such as granulitic breccias which are more prevalent lower in the massifs and which we infer to be older (pre-basin) highland components. (3) Soil from station 6, North Massif, contains magnesian troctolitic anorthosite, which is a component that is rare in station-2 South-Massif,contains magnesian troctolitic in impact-melt breccia interpreted by most investigators to be ejecta from the Serenitatis basin.

Jolliff, Bradley L.↗

DOE OSTI2025

End-to-end deep learning pipeline for real-time Bragg peak segmentation: from training to large-scale deployment

X-ray crystallography reconstruction, which transforms discrete X-ray diffraction patterns into three-dimensional molecular structures, relies critically on accurate Bragg peak finding for structure determination. As X-ray free electron laser (XFEL) facilities advance toward MHz data rates (1 million images per second), traditional peak finding algorithms that require manual parameter tuning or exhaustive grid searches across multiple experiments become increasingly impractical. While deep learning approaches offer promising solutions, their deployment in high-throughput environments presents significant challenges in automated dataset labeling, model scalability, edge deployment efficiency, and distributed inference capabilities. We present an end-to-end deep learning pipeline with three key components: (1) a data engine that combines traditional algorithms with our peak matching algorithm to generate high-quality training data at scale, (2) a modular architecture that scales from a few million to hundreds of million parameters, enabling us to train large expert-level models offline while deploying smaller, distilled models at the edge, and (3) a decoupled producer-consumer architecture that separates specialized data source layer from model inference, enabling flexible deployment across diverse computing environments. Using this integrated approach, our pipeline achieves accuracy comparable to traditional methods tuned by human experts while eliminating the need for experiment-specific parameter tuning. Although current throughput requires optimization for MHz facilities, our system's scalable architecture and demonstrated model compression capabilities provide a foundation for future high-throughput XFEL deployments.

Wang, Cong↗

NASA NTRS1972

A computer program to evaluate optical systems

A computer program is used to evaluate a 25.4 cm X-ray telescope at a field angle of 20 minutes of arc by geometrical analysis. The object is regarded as a point source of electromagnetic radiation, and the optical surfaces are treated as boundary conditions in the solution of the electromagnetic wave propagation equation. The electric field distribution is then determined in the region of the image and the intensity distribution inferred. A comparison of wave analysis results and photographs taken through the telescope shows excellent agreement.

Innes, D.↗

NASA NTRS1975

Molecular hydrogen in the Galaxy and galactic gamma rays

Recent surveys of 2.6-cm CO emission and 100-MeV gamma-radiation in the galactic plane reveal a striking correlation suggesting that both emissions may be primarily proportional to the line-of-sight column density of H2 in the inner Galaxy. Both the gamma-ray and CO data suggest a prominent ring or arm consisting of cool clouds of H2 at a galactocentric distance of about 5 kpc with a mean total hydrogen density equivalent to approximately 5 atoms per cu cm. Estimates are made of column densities of H2 at 0 deg galactic longitude and are compared with estimates from infrared and X-ray absorption measurements. These estimates are all consistent, indicating that H2 is far more abundant than H I in the inner Galaxy and is the key to a more satisfactory explanation of the gamma-ray observations than previous suggestions. The importance of H2 in understanding galactic gamma-ray observations is also reflected in the correlation of galactic-latitude distribution of gamma-rays and dense dust clouds. The deduced cosmic-ray distribution inferred from the calculations is similar to that of galactic supernova remnants, suggesting a galactic origin for most cosmic rays.

Stecker, F. W.↗

NASA NTRS1976

Coincident observations of ionospheric troughs and the equatorial plasmapause

Electron-density observations made in the topside ionosphere by the Ariel 4 and Isis 2 satellites are examined in conjunction with results obtained by Explorer 45 when it traversed the near-equatorial plasmapause with one hour (both UT and MLT) of the Ariel and Isis traversals of the same L coordinate. Both dusk and night observations are analyzed, and an attempt is made to show that depressions in ionospheric electron density occur in the vicinity of the plasmapause field line. It is concluded that the electron distributions observed in the electron-density troughs at 550 km near dusk by Ariel and at 1400 km near midnight by Isis do not always parallel variations in the light-ion distribution inferred from the Explorer plasmapause traversals and that there appears to be no specific feature of the main ionospheric trough which can be used to identify the plasmapause field line except in a statistical sense.

Grebowsky, J. M.↗