Search NASA⌕ Search

SEARCH · Search NASA

Results for “scientific method”

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 325 records · Page 18
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

UZrCN Synthesis via Arc Melting

The next generation of nuclear reactors for both power production and space nuclear propulsion require fuel that is more durable, thermally stable, and more thermally conductive to support rapid heat transfer. High temperature gas reactors (HTGR), advanced gas reactors (AGR), and space-based nuclear thermal propulsion (NTP) are advanced reactor concepts that require a fuel type that can withstand high temperatures (1000-2900K) and flow of corrosive gas coolants such as helium, hydrogen, and carbon dioxide. One fuel with the potential to meet these demanding requirements is uranium-zirconium-carbonitride (UZrCN). UZrCN has many favorable fuel qualities compared to other eligible fuel forms such as uranium dioxide (UO 2 ) and uranium mononitride (UN) that could support the aforementioned reactor concepts. UZrCN has an exceptionally high operating temperature and thermal conductivity which are highly desirable to improve reactor economics and safety. It far exceeds the properties of UO 2 which is the most common fuel form in the United States. UZrCN also surpasses UN in terms of thermal conductivity and operating temperature by eliminating the dissociation problem UN has at 1700K. UZrCN could improve gas reactor performance and enable NTP technologies; however, it is an under-researched fuel that lacks rigorous scientific study. In recent efforts by the Idaho National Laboratory, a variety of novel methods to produce this fuel composition have been explored. One such method is via arc melting of uranium, zirconium, and carbon under a nitrogen atmosphere. Alloy fabrication using arc melting has been utilized for close to 150 years now and is well-understood as a method for rapid alloy prototyping. This process will be used to perform in-situ nitriding to form UZrCN.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2021

Electrochemical Manipulation and Radiolytic Evaluation of Organic Phase Neptunium

Presentation for Seed LDRD proposal presentation: This Seed LDRD research proposes to test an innovative approach to precisely control the oxidation state distribution of neptunium (Np) under used nuclear fuel (UNF) reprocessing conditions using novel electrochemical methods. Current UNF reprocessing technologies are not financially viable in the US, requiring significant scientific and technological innovation to improve cost efficiency. Reducing the number of process cycles is one avenue for increasing cost efficiency, of which resolving the challenge associated with isolating Np is critical. Under envisioned process conditions, Np is present in a mixture of extractable (Np(IV) and Np(VI)) and inextractable (Np(V)) oxidation states, the distribution of which is dependent on several factors, including ionizing radiation dose, that change throughout a reprocessing scheme. Consequently, Np unintentionally partitions into various UNF reprocessing phases and product streams, ultimately increasing the number of process cycles to isolate UNF components. Here, we propose to employ novel, high surface area, optically transparent, Ligand Modified tin-doped indium oxide Electrodes (LMES). These LMEs bind Np, and therefore enable concurrent generation and spectroscopic characterization of Np oxidation states in organic solutions. These proof-of-concept experiments will: (i) facilitate optimization of the proposed electrochemical system conditions for Np oxidation state manipulation; (ii) identify the accessible electrochemical window for Np complexes in the organic phase; (iii) report characteristic optical spectra for each accessible complexed Np oxidation state; and (iv) determine the lifetime and partitioning of a given atypical Np oxidation state in the organic phase in the presence and absence of ionizing radiation fields. The data gathered by this Seed LDRD will provide support for the design of an electrochemical process concept for the precise manipulation of Np oxidation states in UNF reprocessing solvent systems, with the intention of providing advanced control over Np mass transfer, and thus greater process efficiency and economy.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2024

UZrCN Synthesis via Arc Melting - A Novel Synthesis Study

The next generation of nuclear reactors for both power production and space nuclear propulsion require fuel that is more durable, thermally stable, and more thermally conductive to support rapid heat transfer. High temperature gas reactors (HTGR), advanced gas reactors (AGR), and space-based nuclear thermal propulsion (NTP) are advanced reactor concepts that require a fuel type that can withstand high temperatures (1000-2900K) and flow of corrosive gas coolants such as helium, hydrogen, and carbon dioxide. One fuel with the potential to meet these demanding requirements is uranium-zirconium-carbonitride (UZrCN). UZrCN has many favorable fuel qualities compared to other eligible fuel forms such as uranium dioxide (UO2) and uranium mononitride (UN) that could support the aforementioned reactor concepts. UZrCN has an exceptionally high operating temperature and thermal conductivity which are highly desirable to improve reactor economics and safety. It far exceeds the properties of UO2 which is the most common fuel form in the United States. UZrCN also surpasses UN in terms of thermal conductivity and operating temperature by eliminating the dissociation problem UN has at 1700K. UZrCN could improve gas reactor performance and enable NTP technologies; however, it is an under-researched fuel that lacks rigorous scientific study. In recent efforts by the Idaho National Laboratory, a variety of novel methods to produce this fuel composition have been explored. One such method is via arc melting of uranium, zirconium, and carbon under a nitrogen atmosphere. Alloy fabrication using arc melting has been utilized for close to 150 years now and is well-understood as a method for rapid alloy prototyping. This process will be used to perform in-situ nitriding to form UZrCN.

36 MATERIALS SCIENCE↗
DOE OSTI2024

UZrCN Formation via Arc Melting – A Novel Synthesis Study

The next generation of nuclear reactors for both power production and space nuclear propulsion require fuel that is more durable, thermally stable, and more thermally conductive to support rapid heat transfer. High temperature gas reactors (HTGR), advanced gas reactors (AGR), and space-based nuclear thermal propulsion (NTP) are advanced reactor concepts that require a fuel type that can withstand high temperatures (1000-2900K) and flow of corrosive gas coolants such as helium, hydrogen, and carbon dioxide. One fuel with the potential to meet these demanding requirements is uranium-zirconium-carbonitride (UZrCN). UZrCN has many favorable fuel qualities compared to other eligible fuel forms such as uranium dioxide (UO2) and uranium mononitride (UN) that could support the aforementioned reactor concepts. UZrCN has an exceptionally high operating temperature and thermal conductivity which are highly desirable to improve reactor economics and safety. It far exceeds the properties of UO2 which is the most common fuel form in the United States. UZrCN also surpasses UN in terms of thermal conductivity and operating temperature by eliminating the dissociation problem UN has at 1700K. UZrCN could improve gas reactor performance and enable NTP technologies; however, it is an under-researched fuel that lacks rigorous scientific study. In recent efforts by the Idaho National Laboratory, a variety of novel methods to produce this fuel composition have been explored. One such method is via arc melting of uranium, zirconium, and carbon under a nitrogen atmosphere. Alloy fabrication using arc melting has been utilized for close to 150 years now and is well-understood as a method for rapid alloy prototyping. This process will be used to perform in-situ nitriding to form UZrCN.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Intern-Artificial Intelligence Benchmarking

Benchmarks provide a standardized method for evaluating different AI models, enabling reproducibility and comparison between models, and facilitating scientific progress. As AI models continue to develop rapidly, incorporating new datasets, capabilities, and architectures becomes more complicated. Therefore, the current static benchmarks become increasingly irrelevant. The MLCommons team argues that to make AI benchmarks more relevant, it involves making the benchmarks themselves more dynamic, as well as technical innovations that make it easier for scientists and researchers at all levels to use and contribute to the benchmarks. The current progress in technical innovation is a software that allows for a detailed view of a collection of AI benchmarks to be output in various formats that are easily readable and accessible.

Krishnan, Anjay [Fermilab]↗
DOE OSTI2024

Project DarkStar: Vision for LLNL in 2030

DarkStar was a Strategic Initiative (FY2021-FY2024) to investigate applications of Artificial Intelligence (AI) and Machine Learning (ML) to scientific problems of complex hydrodynamics, shockwave physics and energetic materials. The research focused on physics and engineering design as a process that can be tremendously accelerated through merging AI with advanced physics simulation on exascale-class platforms, and to experimentally validate this revolutionary new approach through dynamic materials campaigns. A central thread of scientific inquiry was in the application of AI to enable human understanding of how to control hydrodynamic instability (which has impacts to areas such as inertial confinement fusion) via engineering features and time-dependent sources. Motivated by an unfinished line of research started by Dr. Johnny von Neumann, AI-enabled simulation approaches were developed that allowed DarkStar researchers to uncover several ground-breaking discoveries regarding hydrodynamic instability, including how to completely suppress Richtmyer-Meshkov instability (RMI). These S&T discoveries, along with other advances, have shown the way for an entirely new approach to time-dependent problems known as inverse design – the idea that complex systems can be developed directly from a final state that is to be achieved and resolve the initial design via satisfying several constraints simultaneously via AI/ML. Through experimental campaigns conducted across a wide range of facilities in the NNSA complex (the High Explosive Application Facility at LLNL, the Dynamic Compression Sector/Advanced Photon Source at Argonne National Lab, and Special Technologies Laboratory at MSTS) the radical new AI/ML approach to engineering complex material dynamics was verified, establishing a new field of study within the realm of shock physics. As advanced manufacturing capabilities continue to develop, the great importance of inverse design as a means to apply that technology effectively for NNSA missions will feature prominently over this decade. DarkStar has positioned NNSA as a world-leader in this newly emerging cross-disciplinary area of AI methods for advanced physics simulation and pioneered multiple novel approaches that have enabled the broader scientific community. By allowing us to see past the horizon, to 2030 and beyond, DarkStar has illuminated the vast potential of AI/ML to impact a wide range of new national security missions and, consequently, multiple areas of further research have already emerged across the NNSA and DOD complex.

42 ENGINEERING↗
DOE OSTI2022

Durability Evaluation of Advanced Fenestration Technologies

In the U.S., more than 40% of primary energy and 70% of electricity is consumed in residential and commercial buildings, resulting in annual energy costs of more than $430 billion. Approximately 35% of this consumption can be attributed to losses through the building envelope, of which windows are currently the weakest link. Multiple technologies are under development to improve this performance. This includes dynamic and photovoltaic glazing as well as emerging highly insulating technologies including vacuum insulating glass (VIG), aerogels and thin-glass based multi-pane configurations. While windows are specified based on expected performance as installed, the energy savings realized by high performance windows are delivered over time. This makes it critical to understand and maximize the durability of high-performance windows to ensure those projected energy savings are delivered. Present methods for evaluating durability are based on existing technologies. These methods may not apply or be adequate for newer emerging technologies which often present novel failure mechanisms that need to be understood and evaluated differently. In this presentation, we will discuss our efforts to define appropriate methods to improve the evaluation of many existing technologies as well as evaluate the durability of emerging window technologies. This work results from the combination of an extensive review of various international standards as well as existing scientific literature. In addition, input was gathered from multiple industry stakeholders regarding present practice as well as feedback on proposed improvements to existing methods. Here we will share these findings and present proposed improvements to developing and existing durability evaluation protocols.

building envelope↗
DOE OSTI2020

Experiences with a Flexible User Research Process to Build Data Change Tools

Scientific software development processes are understood to be distinct from commercial software development practices due to uncertain and evolving states of scientific knowledge. Sustaining these software products is a recognized challenge, but under-examined is the usability and usefulness of such tools to their scientific end users. User research is a well-established set of techniques (e.g., interviews, mockups, usability tests) applied in commercial software projects to develop foundational, generative, and evaluative insights about products and the people who use them. Currently these approaches are not commonly applied and discussed in scientific software development work. The use of user research techniques in scientific environments can be challenging due to the nascent, fluid problem spaces of scientific work, varying scope of projects and their user communities, and funding/economic constraints on projects.In this paper, we reflect on our experiences undertaking a multi-method user research process in the Deduce project. The Deduce project is investigating data change to develop metrics, methods, and tools that will help scientists make decisions around data change. There is a lack of common terminology since the concept of systematically measuring and managing data change is under explored in scientific environments. To bridge this gap we conducted user research that focuses on user practices, needs, and motivations to help us design and develop metrics and tools for data change. This paper contributes reflections and the lessons we have learned from our experiences. We offer key takeaways for scientific software project teams to effectively and flexibly incorporate similar processes into their projects.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Data-Enabled Fusion Technology (Final Scientific/Technical Report)

Advancing Scientific Understanding in Fusion Energy and Machine Learning This research represented a significant step forward in machine learning (ML) applications for fusion energy experiments. The project integrated advanced data-driven modeling, optimization techniques, and artificial intelligence to enhance the predictive capabilities and operational efficiency of plasma-based fusion systems. Specifically, tasks focused on ML-enhanced diagnostics, operator guidance tools, and predictive modeling helped improve the ability to interpret complex fusion experiments. Key areas of advancement included: 1) data-driven plasma control, i.e., using ML algorithms to optimize experimental conditions and classify plasma behaviors based on historical data; 2) spectroscopy and diagnostics, i.e., applying AI models to extract previously inaccessible insights from experimental spectroscopy data; and 3) configuration mapping and operator guidance, i.e., developing a predictive framework to assist scientists in identifying the most effective experimental parameters, reducing reliance on manual adjustments. By refining these ML-driven techniques, the project contributed to the broader scientific community’s understanding of plasma dynamics and fusion energy viability. Technical Effectiveness and Economic Feasibility The methods investigated demonstrated high technical effectiveness, as reflected in milestones assessing the predictive accuracy, performance, and optimization of fusion configurations. The development of an Operator Guidance Tool (OGT), for example, led to more precise control of plasma conditions by learning from experimental data and offering real-time adjustments. From an economic standpoint, DeFT provided: 1) the ability to reduce trial-and-error experimentation, which lowered operational costs; 2) improved data interpretation methods, which enabled more efficient resource allocation in large-scale fusion research projects; and 3) the automation of key diagnostic tasks, which reduced manual labor and human error, increasing overall efficiency. 13 The final assessments of predictive models and optimization strategies demonstrated that these approaches were scalable and could be implemented across multiple fusion energy research programs. Public Benefit and Societal Impact This project contributed directly to the broader goal of achieving sustainable and commercially viable fusion energy, which had profound implications for clean energy production and climate change mitigation. The integration of AI-driven solutions into fusion research: 1) sped up scientific discovery, accelerating progress towards achieving energy breakthroughs; 2) reduced the cost of experimentation, making fusion research more accessible; and 3) provided a framework for future AI applications in high-energy physics, benefiting adjacent fields like space exploration, material science, and renewable energy. Additionally, by fostering collaborations between AI researchers and plasma physicists, this project promoted interdisciplinary innovation that could lead to broader applications beyond fusion research.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

Total metals & anion concentration data; Slate River floodplain, Crested Butte, CO; May 2020-September 2020

This data package includes processed and undiluted measurements for metal and anion concentrations from pore water (groundwater) samples from the Slate River floodplain of Crested Butte, CO, a focus field site for the SLAC Floodplain Hydro-Biogeochemistry SFA. The data was generated as part of the work targeting the overarching research question for the SLAC SFA: How do ubiquitous subsurface interfaces mediate molecular-scale biogeochemical processes and groundwater quality in floodplains and watersheds? Samples were collected between May and September of 2020. These measurements were all recorded at the Arizona Laboratory for Emerging Contaminants (ALEC) at the University of Arizona located in Tucson, AZ. Groundwater samples were extracted from a network of installed rhizon (Rhizosphere Research Products, part no. 19.60.21F, 0.6 micrometer mesh size) and piezometer wells within the river floodplain. All water samples were shaded from sun exposure during extraction from the subsurface and preserved at 4C until measured at ALEC.Analysis by ICP-MS:Measurements for total metals were made on the Agilent 7700x ICP-MS (for total metals) – Agilent Technologies, Santa Clara, CA.The analytical QA/QC protocol was adapted from US EPA Method 200.8 for analysis by ICP-MS. Calibration standards were prepared from multi-element stock solution (Sigma-Aldrich Multielement standard solution for ICP, St. Louis, MO) using matrix matched to sample solutions (either 2% HCl or HNO3 from AriStar Plus,grade acids from VWR Scientific). Calibration curves include at least 7 points with correlation coefficients > 0.995. The QC protocol includes a continuing calibration blank (CCB), a continuing calibration verification (CCV) solution and at least one quality control sample (QCS) to be analyzed just after calibration and again after every 12 samples and at the completion of the run. The QCS solutions are from an independent source, such as NIST SRM 1643e - Trace Elements in Water, or QCS solutions from High Purity Standards (Charleston, SC). Acceptable QC responses must be between 90 and 110% of the certified value. An internal standard (Rh) is added via on-line addition into the sample line using a mixing tee.Analysis by Ion Chromatography (Anions):The protocol follows Method 4110 in Standard Methods for Examination of Water and Wastewater.The instrument used is the Thermo Scientific Dionex ICS-6000 using AS+AG22 column set for anion analysis with isocratic method using sodium carbonate eluent. Detection is by chemical suppression of eluent conductivity. Quality control solutions and mixed analyte standards purchased from Inorganic Ventures, Christiansburg, VA.All files are in csv format.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Foundations of machine learning for low-temperature plasmas: methods and case studies

Abstract Machine learning (ML) and artificial intelligence have proven to be an invaluable tool in tackling a vast array of scientific, engineering, and societal problems. The main drivers behind the recent proliferation of ML in practically all aspects of science and technology can be attributed to: (a) improved data acquisition and inexpensive data storage; (b) exponential growth in computing power; and (c) availability of open-source software and resources that have made the use of state-of-the-art ML algorithms widely accessible. The impact of ML on the field of low-temperature plasmas (LTPs) could be particularly significant in the emerging applications that involve plasma treatment of complex interfaces in areas ranging from the manufacture of microelectronics and processing of quantum materials, to the LTP-driven electrification of the chemical industry, and to medicine and biotechnology. This is primarily due to the complex and poorly-understood nature of the plasma-surface interactions in these applications that pose unique challenges to the modeling, diagnostics, and predictive control of LTPs. As the use of ML is becoming more prevalent, it is increasingly paramount for the LTP community to be able to critically analyze and assess the concepts and techniques behind data-driven approaches. To this end, the goal of this paper is to provide a tutorial overview of some of the widely-used ML methods that can be useful, amongst others, for discovering and correlating patterns in the data that may be otherwise impractical to decipher by human intuition alone, for learning multivariable nonlinear data-driven prediction models that are capable of describing the complex behavior of plasma interacting with interfaces, and for guiding the design of experiments to explore the parameter space of plasma-assisted processes in a systematic and resource-efficient manner. We illustrate the utility of various supervised, unsupervised and active learning methods using LTP datasets consisting of commonly-available, information-rich measurements (e.g. optical emission spectra, current–voltage characteristics, scanning electron microscope images, infrared surface temperature measurements, Fourier transform infrared spectra). All the ML demonstrations presented in this paper are carried out using open-source software; the datasets and codes are made publicly available. The FAIR guiding principles for scientific data management and stewardship can accelerate the adoption and development of ML in the LTP community.

Physics↗
DOE OSTI2024

Accuracy versus precision in boosted top tagging with the ATLAS detector

The identification of top quark decays where the top quark has a large momentum transverse to the beam axis, known as top tagging , is a crucial component in many measurements of Standard Model processes and searches for beyond the Standard Model physics at the Large Hadron Collider. Machine learning techniques have improved the performance of top tagging algorithms, but the size of the systematic uncertainties for all proposed algorithms has not been systematically studied. This paper presents the performance of several machine learning based top tagging algorithms on a dataset constructed from simulated proton-proton collision events measured with the ATLAS detector at $\sqrt{s}$ = 13 TeV. The systematic uncertainties associated with these algorithms are estimated through an approximate procedure that is not meant to be used in a physics analysis, but is appropriate for the level of precision required for this study. The most performant algorithms are found to have the largest uncertainties, motivating the development of methods to reduce these uncertainties without compromising performance. To enable such efforts in the wider scientific community, the datasets used in this paper are made publicly available.

47 OTHER INSTRUMENTATION↗
DOE OSTI2020

The Future of Computing: Integrating Scientific Computation on Neuromorphic Systems

Neuromorphic computing is known for its integration of algorithms and hardware elements that are inspired by the brain. Conventionally, this nontraditional method of computing is used for many neural or learning inspired applications. Unfortunately, this has resulted in the field of neuromorphic computing being relatively narrow in scope. In this paper we discuss two research areas actively trying to widen the impact of neuromorphic systems. The first is Fugu, a high-level programming interface designed to bridge the gap between general computer scientists and those who specialize in neuromorphic areas. The second aims to map classical scientific computing problems onto these frameworks through the example of random walks. This elucidates a class of scientific applications that are conducive to neuromorphic algorithms.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Hutchinson Trace Estimation for high-dimensional and high-order Physics-Informed Neural Networks

Physics-Informed Neural Networks (PINNs) have proven effective in solving partial differential equations (PDEs), especially when some data are available by seamlessly blending data and physics. However, extending PINNs to high-dimensional and even high-order PDEs encounters significant challenges due to the computational cost associated with automatic differentiation in the residual loss function calculation. Herein, we address the limitations of PINNs in handling high-dimensional and high-order PDEs by introducing the Hutchinson Trace Estimation (HTE) method. Starting with the second-order high-dimensional PDEs, which are ubiquitous in scientific computing, HTE is applied to transform the calculation of the entire Hessian matrix into a Hessian vector product (HVP). This approach not only alleviates the computational bottleneck via Taylor-mode automatic differentiation but also significantly reduces memory consumption from the Hessian matrix to an HVP’s scalar output. We further showcase HTE’s convergence to the original PINN loss and its unbiased behavior under specific conditions. Comparisons with the Stochastic Dimension Gradient Descent (SDGD) highlight the distinct advantages of HTE, particularly in scenarios with significant variability and variance among dimensions. We further extend the application of HTE to higher-order and higher-dimensional PDEs, specifically addressing the biharmonic equation. By employing tensor-vector products (TVP), HTE efficiently computes the colossal tensor associated with the fourth-order high-dimensional biharmonic equation, saving memory and enabling rapid computation. The effectiveness of HTE is illustrated through experimental setups, demonstrating comparable convergence rates with SDGD under memory and speed constraints. Additionally, HTE proves valuable in accelerating the Gradient-Enhanced PINN (gPINN) version as well as the Biharmonic equation. Overall, HTE opens up a new capability in scientific machine learning for tackling high-order and high-dimensional PDEs.

Curse of dimensionality↗
DOE OSTI2021

An Introduction to the Benchmarking and Publications for Non-Targeted Analysis Working Group

Non-targeted analysis (NTA) encompasses a rapidly evolving set of mass spectrometry techniques aimed at characterizing the chemical composition of complex samples, identifying unknown compounds, and/or classifying samples, without prior knowledge regarding the chemical content of the samples. Recent advances in NTA are the result of improved and more accessible instrumentation for data generation, and analysis tools for data evaluation and interpretation. As researchers continue to develop NTA approaches in various scientific fields, there is a growing need to identify, disseminate, and adopt community-wide method reporting guidelines. In 2018, NTA researchers formed the Benchmarking and Publications for Non-Targeted Analysis Working Group (BP4NTA) to address this need. Consisting of participants from around the world and representing fields ranging from environmental science and food chemistry to ‘omics and toxicology, BP4NTA provides resources addressing a variety of challenges associated with NTA. Furthermore, thus far, BP4NTA group members have aimed to establish consensus on NTA-related terms and concepts, and to create consistency in reporting practices by providing resources on a public website, including consensus definitions, reference content, and lists of available tools. Moving forward, BP4NTA will provide a setting for NTA researchers to continue discussing emerging challenges and contribute to additional harmonization efforts.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2026

A variational framework for residual-based adaptivity in neural PDE solvers and operator learning

Residual-based adaptive strategies are widely used in scientific machine learning yet remain largely heuristic. We introduce a variational framework that formalizes these methods through convex transformations of the residual, where different transformations correspond to distinct objective functionals. For instance, exponential weights target uniform error minimization, while linear weights recover quadratic error minimization. This perspective reveals adaptive weighting as a means of selecting sampling distributions that optimize a primal objective, directly linking discretization choices to error metrics. This principled approach yields three key benefits: it enables systematic design of adaptive schemes, reduces discretization error by lowering estimator variance, and enhances learning dynamics by improving gradient signal-to-noise ratio. Extending the framework to operator learning, we demonstrate substantial performance gains across diverse optimizers and architectures. Our results provide a theoretical perspective for residual-based adaptivity and establish a foundation for principled discretization and training.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Constrained GAN-Generated X-Ray CT Data For Self-Supervised And Foundation-Model Segmentation Of Concrete Microstructures

Three-dimensional characterization of materials using X-ray computed tomography (XCT) is challenging due to the complexity of internal structures, noise, and variations in resolution. Traditional computer vision models often struggle to accurately segment these images, particularly in domain-specific applications like materials science. While supervised deep learning approaches have been developed to address the limitations of conventional algorithms, they typically require large amounts of labeled training data and often fail to generalize across different datasets. Self-supervised, few-and zero-shot learning methods have gained prominence in natural image processing and segmentation tasks, but their application to scientific imaging remains limited due to the unique structural complexity, noise, and textural artifacts present in materials science data. In this work, we investigate how domain adaptation, leveraging physics-based and GAN-generated synthetic data, impacts segmentation performance. We introduce a modified Contrastive Unpaired Translation (CUT) model designed to generate realistic labeled data, which can be used for training, pre-training, and fine-tuning segmentation models for real XCT microstructure data. We evaluate the performance of two segmentation approaches: a self-supervised network (SSL-ALPNet) and a foundation model (Segment Anything Model), assessing their improvements when pre-trained and/or fine-tuned on the synthesized data. Our results demonstrate that leveraging synthetic data significantly enhances segmentation performance, particularly in challenging materials science applications.

Ziabari, Amir [ORNL] (ORCID:000000034776457X)↗
DOE OSTI2024

GPU Accelerated Sparse Cholesky Factorization

The solution of sparse symmetric positive definite linear systems is an important computational kernel in large-scale scientific and engineering modeling and simulation. We will solve the linear systems using a direct method, in which a Cholesky factorization of the coefficient matrix is performed using a right-looking approach and the resulting triangular factors are used to compute the solution. Sparse Cholesky factorization is compute intensive. In this work we investigate techniques for reducing the factorization time in sparse Cholesky factorization by offloading some of the dense matrix operations on a GPU. We will describe the techniques we have considered. We achieved up to 4x speedup compared to the CPU-only version.

Karsavuran, M Ozan↗