Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Visual Examination”

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 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Geometric Interpretation of the Cluster Location Problem Part I: Theory

We present a new framing of the seismic location problem using principles drawn from differential geometry. Our interpretation relies upon the common assumption that travel times observed across a network are continuous, differentiable functions of source location. In consequence, travel‐time functions constitute a differentiable map between the source region and a Riemannian manifold. The manifold is said to be the image of the source region embedded in a generally high‐dimension travel‐time vector space. A cluster of events in the source region has an image of discrete points on the manifold, that, except in the simplest cases, cannot be viewed directly. However, it is possible to project the image of a cluster into a tangent space of the manifold for direct visualization. The projection operator can be computed directly from the data without a velocity model, but produces a distorted rendering of the cluster geometry. With a model we can predict the distortions and correct them to estimate cluster geometry. We develop these points with the simplest possible example, one for which direct visualization of the manifold is possible, using the example as an introduction to the relevant concepts from differential geometry in a familiar setting. The tangent space, a local linearization of the manifold, plays a key role. We develop a metric to estimate the limits of linearization, that is, to determine when the curvature of the manifold invalidates the linear assumption. We also examine the interplay of model error, inadequate network geometry, and pick error. We then generalize our results from the simple case to the general case of 3D source regions observed by general networks. Although we do suggest a new “project and correct” method for location, we do not develop it into a practical algorithm. In conclusion, our intention rather is to highlight new analytical methods grounded in differential geometry.

East Pacific Ocean Islands↗
DOE OSTI2026

An uncertainty visualization framework for large-scale cardiovascular flow simulations: A case study on aortic stenosis

We present a generalizable uncertainty quantification (UQ) and visualization framework for lattice Boltzmann method simulations of high Reynolds number vascular flows, demonstrated on a patient-specific stenosed aorta. The framework combines EasyVVUQ for parameter sampling with large-eddy simulation turbulence modeling in HemeLB, and executes ensembles on the Frontier exascale supercomputer. Spatially resolved metrics, including entropy and isosurface-crossing probability, are used to map uncertainty in pressure and wall shear stress fields directly onto vascular geometries. Two sources of model variability are examined: inlet peak velocity and the Smagorinsky constant. Inlet velocity variation produces high uncertainty downstream of the stenosis where turbulence develops, while upstream regions remain stable. Smagorinsky constant variation has little effect on the large-scale pressure field but increases WSS uncertainty in localized high-shear regions. In both cases, the stenotic throat manifests low entropy, indicative of robust identification of elevated WSS. By linking quantitative UQ measures to three-dimensional anatomy, the framework improves interpretability over conventional 1D UQ plots and supports clinically relevant decision-making, with broad applicability to vascular flow problems requiring both accuracy and spatial insight.

Hemodynamics↗
DOE OSTI2021

Fractal dimension of optical cirrus in Stripe82

ABSTRACT The geometric characteristics of dust clouds provide important information on the physical processes that structure such clouds. One of such characteristics is the 2D fractal dimension D of a cloud projected on to the sky plane. In previous studies, which were mostly based on infrared (IR) data, the fractal dimension of individual clouds was found to be in a range from 1.1 to 1.7 with a preferred value of 1.2–1.4. In this work, we use data from Stripe82 of the Sloan Digital Sky Survey to measure the fractal dimension of the cirrus clouds. This is done here for the first time for optical data with significantly better resolution as compared to IR data. To determine the fractal dimension, the perimeter-area method is employed. We also consider IR (IRAS and Herschel) counterparts of the corresponding optical fields to compare the results between the optical and IR. We find that the averaged fractal dimension across all clouds in the optical is $\langle D \rangle =1.69^{+0.05}_{-0.05}$ which is significantly larger than the fractal dimension of its IR counterparts $\langle D\rangle =1.38^{+0.07}_{-0.06}$. We examine several reasons for this discrepancy (choice of masking and minimal contour level, image and angular resolution, etc.) and find that for approximately half of our fields the different angular resolution (point spread function) of the optical and IR data can explain the difference between the corresponding fractal dimensions. For the other half of the fields, the fractal dimensions of the IR and visual data remain inconsistent, which can be associated with physical properties of the clouds, but further physical simulations are required to prove it.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Volcano infrasound: progress and future directions

Over the past two decades (2000–2020), volcano infrasound (acoustic waves with frequencies less than 20 Hz propagating in the atmosphere) has evolved from an area of academic research to a useful monitoring tool. As a result, infrasound is routinely used by volcano observatories around the world to detect, locate, and characterize volcanic activity. It is particularly useful in confirming subaerial activity and monitoring remote eruptions, and it has shown promise in forecasting paroxysmal activity at open-vent systems. Fundamental research on volcano infrasound is providing substantial new insights on eruption dynamics and volcanic processes and will continue to do so over the next decade. The increased availability of infrasound sensors will expand observations of varied eruption styles, and the associated increase in data volume will make machine learning workflows more feasible. More sophisticated modeling will be applied to examine infrasound source and propagation effects from local to global distances, leading to improved infrasound-derived estimates of eruption properties. Future work will use infrasound to detect, locate, and characterize moving flows, such as pyroclastic density currents, lahars, rockfalls, lava flows, and avalanches. Infrasound observations will be further integrated with other data streams, such as seismic, ground- and satellite-based thermal and visual imagery, geodetic, lightning, and gas data. The volcano infrasound community should continue efforts to make data and codes accessible and to improve diversity, equity, and inclusion in the field. In summary, the next decade of volcano infrasound research will continue to advance our understanding of complex volcano processes through increased data availability, sensor technologies, enhanced modeling capabilities, and novel data analysis methods that will improve hazard detection and mitigation.

58 GEOSCIENCES↗
DOE OSTI2023

Surface Properties of Double-Fin Generated Shock-Wave/Boundary-Layer Interactions

Shock-Wave/Boundary-Layer Interaction (SBLI) is ubiquitous in high-speed air vehicles. SBLI surface flowfield properties of crossing shocks induced by symmetric Double-Fins (DFs) at fin angles of 8° and 10° at Mach 2 are experimentally investigated; a 10° Single-Fin (SF) is also explored as a comparison. Surface visualization is used to capture distinct flowfield structures such as separation and upstream influence. Steady and unsteady Pressure-Sensitive Paints (PSPs) are used to obtain global surface pressure fields. Characteristics of DF-SBLI surface topology are discussed by comparing the corresponding SF-SBLI. Here, the effect of fin angle on DF-SBLI is characterized, and unsteady dynamics of the pressure field are examined. Spectral Proper Orthogonal Decomposition (SPOD) is performed on unsteady PSP, and traveling surface pressure waves and their frequency-dependent behavior are discussed. Dispersion relation of surface pressure waves is revealed by performing two-dimensional space–time Fourier transform on unsteady PSP, and the results are compared with the phase velocity distributions obtained from SPOD. Accompanying group velocity behavior is examined for distinct SBLI regions wherein wave propagation is found dependent on its frequency. Steady surface pressures along the centerline are compared with previous experimental data. Fin-tip spacing is found to be an essential parameter for scaling of upstream influence length.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2022

Electrical Analysis of Pulsed Laser Annealed Poly-Si:Ga/SiOx Passivating Contacts

High-efficiency single crystalline silicon (c-Si) solar cells require precise control of dopant diffusion profiles and highly active doping concentrations through thermal annealing. Conventional furnace annealing has been successfully employed for dopant diffusion in aluminum back-surface field (Al-BSF), passivated emitter and rear contact (PERC), and Topcon cells. However, furnace annealing limits some next-generation polycrystalline silicon (poly-Si) on SiOx passivating contacts. This limitation largely affects p-type passivating contacts by not being able to: 1) control the dopant diffusion accurately to prevent B segregation at the SiOx and c-Si interface; and 2) provide highly activated doping concentrations for low contact resistivity. In this contribution, we use a nanosecond excimer laser to examine the passivation quality and electrical performance of both B- and Ga-doped poly-Si/SiOx passivating contacts. The core of this method is to take advantage of the nonequilibrium nature of the anneal through rapid melting and recrystallizing the poly-Si in a short timescale and to achieve doping concentrations above the solid solubility limit. Simulations were performed on polished surfaces to visualize the doping diffusion profiles under different laser conditions, and secondary ion mass spectrometry (SIMS) was used to verify the experimental diffusion profiles post pulsed laser melting (PLM). The results show the dopant diffusion profiles can be tuned precisely through PLM. The electrical analysis using VdP-Hall measurements reveals that the active doping concentration reached 10^21 cm-3 for B and ~2×10^21 cm-3 for Ga in poly-Si, far exceeding their solid solubility limit in Si, with a nearly 100% dopant activation achieved for B, and 20% for Ga. This highly activated dopant profile results in a low contact resistivity of <15 mO·cm^2 for B-doped contacts and <40 mO·cm^2 for Ga-doped contacts. We demonstrate that the passivation quality of the laser annealed contacts shows a linear dependency with the laser energy density and the number of pulses. PLM was also performed on random pyramid textured samples, and we found that preferential melting exists for the tips of pyramids, which can be pursued further for selective pinhole opening. Additionally, we perform thermal stability tests of these hyperdoped passivating contacts and show that under certain laser conditions, the B-doped samples exhibit good thermal stability with near full activation for temperatures up to 800°C. Lastly, we transfer this technique to a larger area by overlapping laser spots and show uniform passivation quality and electrical performance. We are currently implementing this PLM annealed p-type passivating contacts into double-side back-junction poly-Si/SiOx passivating contact devices. Our results show that PLM can not only enable the next generation of high-efficiency c-Si photovoltaics technology but can also benefit the c-Si integrated circuit industry through hyperdoping other semiconductors with atoms that may otherwise have a low solubility.

doping diffusion profile↗
DOE OSTI2024

Color symmetry and confinement as an underlying superconformal structure in holographic QCD

Dedicated to the memory of our colleague, Harald Fritzsch, who, together with Murray Gell-Mann, introduced the color quantum number as the exact symmetry responsible for the strong interaction, thus establishing quantum chromodynamics (QCD) as a fundamental non-Abelian gauge theory. A basic understanding of hadron properties, however, such as confinement and the emergence of a mass scale, from first principles QCD has remained elusive: Hadronic characteristics are not explicit properties of the QCD Lagrangian and perturbative QCD, so successful in the large transverse momentum domain, is not applicable at large distances. In this article, we shall examine how this daunting obstacle is overcome in holographic QCD with the introduction of a superconformal symmetry in anti de Sitter (AdS) space which is responsible for confinement and the introduction of a mass scale within the superconformal group. When mapped to light-front coordinates in physical spacetime, this approach incorporates supersymmetric relations between the Regge trajectories of meson, baryon and tetraquark states which can be visualized in terms of specific SU(3) C color representations of quarks. Finally, we will also briefly discuss here the implications of holographic models for QCD color transparency in view of the present experimental interest.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2026

Real-Space Constrained Density Functional Theory Investigation of Site-Specific, Interfacial Charge Recombination Dynamics Across the Au Nanoparticle/TiO 2 Heterojunction

Au nanoparticle (NP)/TiO 2 heterojunction is a representative system to study interfacial charge transfer in photocatalysis and photovoltaics, where suppressing recombination from TiO 2 to Au can enhance hot carrier extraction. We apply real-space constrained density functional theory (CDFT) with Marcus theory to quantify charge recombination time scales across Au/TiO 2 . This approach enables direct control and visualization of charge-separated states, aligning with site-specific probes like time-resolved X-ray photoelectron spectroscopy (trXPS). We find that the charge-separated state features a bipolaron, with recombination dominated by TiO 2 LUMO to Au HOMO transitions, primarily at interfacial Au sites. Marcus rate predictions are benchmarked with surface hopping methods, quantifying differences in time scales and computational efficiency. Lastly, we examine how the Au cluster size affects the free energy change (ΔG) and reorganization energy (λ), explaining trends in closed-shell systems and highlighting challenges for open-shell extrapolations. Overall, CDFT + Marcus theory provides efficient, mechanistically transparent interfacial charge transfer modeling, and we clearly defined its applicability and limitation.

Glenna, Drew M. [Univ. of Idaho, Idaho Falls, ID (↗
DOE OSTI2021

High-speed synchrotron X-ray phase-contrast imaging for evaluating microscale damage mechanisms and tracking cracking behaviors inside cross-ply GFRCs

In this study, we integrated the high-speed synchrotron X-ray phase-contrast imaging (PCI) with a modified Kolsky compression bar loading platform to visualize the dynamic failure processes of cross-ply glass fiber reinforced composites (GFRCs). Four S-2 glass/SC-15 composite specimens were prepared, having similar thicknesses but different stacking sequences, namely as [0 12 /90 12 ], [90 12 /0 12 ], [0 8 /90 8 /0 8 ], and [90 8 /0 8 /90 8 ]. Three-dimensional synchrotron X-ray computed tomography and scanning electron microscopy (SEM) were employed to examine the microstructures and quantify the fiber volume fractions. Each specimen was notched and subjected to a dynamic three-point flexural loading. The onset of cracking close to the notch tip, crack propagation in 0° or 90° plies and their interface, crack opening, and ultimately failure of the specimen were captured by high-speed synchrotron X-ray PCI. Additional dynamic experiments were performed to determine the average time when the stress wave propagated through the specimen and correlate the X-ray images with the specimen&#x27;s force-deflection response. Finally, the surface morphology of each specimen after the dynamic loading was imaged by SEM. Comparison between real-time X-ray images and post-fracture SEM images demonstrated the capability of the X-ray method to record damage evolution inside composites. Furthermore, the high spatial and temporal resolutions of the X-ray setup and edge enhancement by PCI enabled the identification of microscale damage features within 1 μs. Two damaging processes were identified, crack growth was quantified, and fracture toughness of the composites was evaluated. The method is deemed useful to reveal microscale damage mechanisms and track cracking behaviors inside cross-ply GFRCs under dynamic loading in real time.

42 ENGINEERING↗
DOE OSTI2021

Labile Fe(III) supersaturation controls nucleation and properties of product phases from Fe(II)-catalyzed ferrihydrite transformation

We report Fe(II)-catalyzed ferrihydrite (Fh) transformation to more crystalline iron (oxyhydr)oxide phases is a widely occurring geochemical process which has been extensively studied as a function of Fe(II)/Fh ratios at fixed Fh loadings. However, recent isolation of an intermediate Fe(III) species resulting from Fe(II)-Fh contact that facilitates transformation by dissolution/reprecipitation suggests that the kinetics and properties of product phases will instead depend mostly on its rate of accumulation to a critical concentration, consistent with principles in the classical nucleation theory (CNT). This suggests a dependence both on the loading of Fe(II) on the surface, which controls the rate of labile Fe(III) formation, as well as the available volume of solution, which also impacts how fast it can achieve its critical concentration to nucleate product phases. To specifically examine the latter effect, here we studied transformation of 15 mg Fh in 1 mM FeSO 4 solutions at pH 7.2 in batch suspensions of 30 mL, 150 mL, and 450 mL volumes. Time-dependent concentrations of aqueous Fe(II), surface-associated Fe(II), and resulting labile Fe(III) were monitored along with bulk solids characterization as a function of time. Transmission electron microscopy (TEM) was used to visualize the evolution of phases at identical locations on TEM grids. The collective results show that the rates of Fh loss and emergence of product lepidocrocite (Lp) and goethite (Gt) as well as their phase proportions, nucleation mode and morphological properties depend directly on the rate of accumulation of the labile Fe(III) precursor to its critical concentration, which in our experiments was controlled simply by varying the available volume of solution into which it enters. Statistical analyses of TEM image data suggest that while both heterogeneous and homogeneous nucleation occurred in all experiments, the former was increasingly favored at lower Fh/solution ratio due to its lower nucleation barrier being more favorable at attendant lower supersaturations of Fe(III).

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Microstructural examination of interactions between chromia-based refractory and nuclear glass in a melter

This study shows the effects of nuclear waste glass production on Monofrax K-3 refractory corrosion. A continuously fed research-scale melter containing an Fe- and Ni-rich simulated nuclear waste feed with borosilicate glass-forming chemicals was cyclically melted at 1150°C and idled at 1050°C for a total of 11 weeks. Chemical maps using scanning electron microscopy show the interactions between the quenched melt and the refractory. Nanoscale X-ray-computed tomography was used for a three-dimensional visualization of certain parts of the interface. Unreacted K-3 consists of primarily corundum (Al,Cr) 2 O 3 and spinel (Fe 2+ ,Mg)(Al,Cr) 2 O 4 interlocking crystalline phases. Some of the Cr from the refractory interacts with the Ni and Fe from the melt to form a reaction layer comprising (Ni,Fe 2+ )(Cr,Fe 3+ ) 2 O 4 spinel crystals. Simultaneously, melt components (Na,Si) infiltrate into the refractory. This interaction proceeds at the expense of the integrity of the refractory structure. Intact refractory grains (e.g., (Al,Cr) 2 O 3 ) as well as the reaction layer itself can lose mechanical integrity and spall off into the melt, especially near the top of the melter. As the reaction layer can be a protective boundary for the refractory against further melt infiltration, a reduction in the reaction layer thickness allows an increase in refractory corrosion.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Domain Adaptive Graph Neural Networks for Constraining Cosmological Parameters Across Multiple Data Sets

Deep learning models have been shown to outperform methods that rely on summary statistics, like the power spectrum, in extracting information from complex cosmological data sets. However, due to differences in the subgrid physics implementation and numerical approximations across different simulation suites, models trained on data from one cosmological simulation show a drop in performance when tested on another. Similarly, models trained on any of the simulations would also likely experience a drop in performance when applied to observational data. Training on data from two different suites of the CAMELS hydrodynamic cosmological simulations, we examine the generalization capabilities of Domain Adaptive Graph Neural Networks (DA-GNNs). By utilizing GNNs, we capitalize on their capacity to capture structured scale-free cosmological information from galaxy distributions. Moreover, by including unsupervised domain adaptation via Maximum Mean Discrepancy (MMD), we enable our models to extract domain-invariant features. We demonstrate that DA-GNN achieves higher accuracy and robustness on cross-dataset tasks. Using data visualizations, we show the effects of domain adaptation on proper latent space data alignment. This shows that DA-GNNs are a promising method for extracting domain-independent cosmological information, a vital step toward robust deep learning for real cosmic survey data.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Studying Response to Light in Offices: A Literature Review and Pilot Study

Researchers have been exploring the influence of light on health in office settings for over two decades; however, a greater understanding of physiological responses and technology advancements are shifting the way researchers study the influence of light in realistic environments. New technologies paired with Ecological Momentary Assessments (EMAs) administered via smartphones provide ways to collect information about individual light exposure and occupant response throughout the day. The study aims to document occupant response to tunable lighting in a real office environment, including potential beneficial or adverse health and well-being effects. Twenty-three office employees agreed to participate in a twelve-week study examining occupant response to two lighting conditions (static vs. dynamic). No significant differences were observed for any of the measures, highlighting the importance and complexity of in-situ studies conducted in realistic environments. While prior office studies have shown a significant influence on daytime sleepiness and sleep quality, research has not shown mood or stress to be significantly impacted by lighting conditions. Correlation analyses regarding lighting satisfaction, environmental satisfaction, and visual comfort demonstrate a significant relationship between certain items that may inform future studies. Further, the high correlation means it is reasonable to assume that many environmental factors in offices can influence occupant behavior and well-being.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

What’s in My Toolkit? A Review of Technologies for Assessing Changes in Habitats Caused by Marine Energy Development

Marine energy devices are installed in highly dynamic environments and have the potential to affect the benthic and pelagic habitats around them. Regulatory bodies often require baseline characterization and/or post-installation monitoring to determine whether changes in these habitats are being observed. However, a great diversity of technologies is available for surveying and sampling marine habitats, and selecting the most suitable instrument to identify and measure changes in habitats at marine energy sites can become a daunting task. We conducted a thorough review of journal articles, survey reports, and grey literature to extract information about the technologies used, the data collection and processing methods, and the performance and effectiveness of these instruments. We examined documents related to marine energy development, offshore wind farms, oil and gas offshore sites, and other marine industries around the world over the last 20 years. A total of 120 different technologies were identified across six main habitat categories: seafloor, sediment, infauna, epifauna, pelagic, and biofouling. The technologies were organized into 12 broad technology classes: acoustic, corer, dredge, grab, hook and line, net and trawl, plate, remote sensing, scrape samples, trap, visual, and others. Visual was the most common and the most diverse technology class, with applications across all six habitat categories. Technologies and sampling methods that are designed for working efficiently in energetic environments have greater success at marine energy sites. In addition, sampling designs and statistical analyses should be carefully thought through to identify differences in faunal assemblages and spatiotemporal changes in habitats.

16 TIDAL AND WAVE POWER↗
DOE OSTI2023

EVs@Scale Next-Gen Profiles - Fleet Utilization 2023

As U.S. fleet operators begin transitioning to electric vehicles (EVs), critical questions arise regarding how to manage this shift without disrupting fleet operations or placing undue stress on the electric grid. A major challenge for fleets is maintaining effective operational schedules while accommodating charging requirements, particularly with high-power charging (HPC) infrastructure, which presents grid stability concerns for utilities. Proposed solutions such as charging substations, megawatt charging systems (MCS), and smart charge management systems (SCMS) offer potential pathways forward, but their effectiveness depends on alignment with real-world fleet behavior and operational constraints. This report investigates the charging and utilization behavior of EV and EVSE fleets actively employing HPC technologies by conducting detailed case study analyses based on telematics data. A suite of predefined metrics—covering charging, routing, and other operational behaviors—is developed to evaluate the impact of fleet activities on grid infrastructure and identify opportunities for optimization. Results highlight variations in charging behavior across fleets, such as weekday versus weekend usage, diurnal charging trends, and the role of operational predictability in enabling SCMS effectiveness. While SCMS can help lower costs and improve energy efficiency for fleets with stable schedules, they may be insufficient for fleets with highly variable or long-haul operations, which may require more robust solutions like MCS. Visualization of aggregated hourly energy metrics reveals that while fleet behaviors are diverse, there are common temporal patterns that could inform infrastructure planning and energy management. These insights emphasize the need for fleet-specific charging strategies that minimize grid impact while supporting reliable fleet operations. Additionally, the report underscores the broader economic stakes of electrification, particularly in high-value markets such as freight, where misaligned transitions could stall EV adoption. By examining current EV and EVSE fleet deployments using predetermined standardized metrics, this study offers a foundation for developing technologies and operational frameworks that support scalable, grid-compatible electrification across a variety of fleet types while establishing a baseline understanding of operational behaviors. In doing so, we aim to ensure that future charging solutions reflect actual fleet needs and grid constraints—an essential step toward maintaining operational continuity and achieving a successful transition to electric fleet operations.

Charging↗
DOE OSTI2022

The relation between quasars’ optical spectra and variability

Abstract Brightness variation is an essential feature of quasars, but its mechanism and relationship to other physical quantities are not understood well. We aimed to find the relationship between the optical variability and spectral features to reveal the regularity behind the random variation. It is known that a quasar’s Fe ii/Hβ flux ratio and equivalent width of [O iii]5007 are negatively correlated; this is called Eigenvector 1. In this work, we visualized the relationship between the position on this Eigenvector 1 (EV1) plane and how the brightness of the quasars had changed after ∼10 yr. We conducted three analyses, using a different quasar sample in each. The first analysis showed the relation between the quasars’ distributions on the EV1 plane and how much they had changed brightness, using 13438 Sloan Digital Sky Survey quasars. This result shows how brightness changes later are clearly related to the position on the EV1 plane. In the second analysis, we plotted the sources reported as “changing-look quasars” (or “changing-state quasars”) on the EV1 plane. This result shows that the position on the EV1 plane corresponds to the activity level of each source, and the bright or dim states of them are distributed on the opposite sides divided by the typical quasar distribution. In the third analysis, we examined the transition vectors on the EV1 plane using sources with multiple-epoch spectra. This result shows that the brightening and dimming sources move on a similar path and they reach a position corresponding to the opposite activity level. We also found this trend is opposite to the empirical rule that $R_{\rm {Fe\, \small {II}}}$ positively correlated with the Eddington ratio, which has been proposed based on the trends of a large number of quasars. From all these analyses, it is indicated that quasars tend to oscillate between both sides of the distribution ridge on the EV1 plane; each of them corresponds to a dim state and a bright state. This trend in optical variation suggests that significant brightness changes, such as changing-look quasars, are expected to repeat.

Astronomy & Astrophysics↗
DOE OSTI2025

I/O in Machine Learning Applications on HPC Systems: A 360-degree Survey

Growing interest in Artificial Intelligence (AI) has resulted in a surge in demand for faster methods of Machine Learning (ML) model training and inference. This demand for speed has prompted the use of high performance computing (HPC) systems that excel in managing distributed workloads. Because data is the main fuel for AI applications, the performance of the storage and I/O subsystem of HPC systems is critical. In the past, HPC applications accessed large portions of data written by simulations or experiments or ingested data for visualizations or analysis tasks. ML workloads perform small reads spread across a large number of random files. This shift of I/O access patterns poses several challenges to modern parallel storage systems. In this paper, we survey I/O in ML applications on HPC systems, and target literature within a 6-year time window from 2019 to 2024. We define the scope of the survey, provide an overview of the common phases of ML, review available profilers and benchmarks, examine the I/O patterns encountered during offline data preparation, training, and inference, and explore I/O optimizations utilized in modern ML frameworks and proposed in recent literature. Lastly, we seek to expose research gaps that could spawn further R&D.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

DeepAdversaries: examining the robustness of deep learning models for galaxy morphology classification

With increased adoption of supervised deep learning methods for work with cosmological survey data, the assessment of data perturbation effects (that can naturally occur in the data processing and analysis pipelines) and the development of methods that increase model robustness are increasingly important. In the context of morphological classification of galaxies, we study the effects of perturbations in imaging data. In particular, we examine the consequences of using neural networks when training on baseline data and testing on perturbed data. We consider perturbations associated with two primary sources: (a) increased observational noise as represented by higher levels of Poisson noise and (b) data processing noise incurred by steps such as image compression or telescope errors as represented by one-pixel adversarial attacks. We also test the efficacy of domain adaptation techniques in mitigating the perturbation-driven errors. We use classification accuracy, latent space visualizations, and latent space distance to assess model robustness in the face of these perturbations. For deep learning models without domain adaptation, we find that processing pixel-level errors easily flip the classification into an incorrect class and that higher observational noise makes the model trained on low-noise data unable to classify galaxy morphologies. On the other hand, we show that training with domain adaptation improves model robustness and mitigates the effects of these perturbations, improving the classification accuracy up to 23% on data with higher observational noise. Domain adaptation also increases up to a factor of ${\approx}2.3$ the latent space distance between the baseline and the incorrectly classified one-pixel perturbed image, making the model more robust to inadvertent perturbations. Successful development and implementation of methods that increase model robustness in astronomical survey pipelines will help pave the way for many more uses of deep learning for astronomy.

79 ASTRONOMY AND ASTROPHYSICS↗