Search NASA⌕ Search

SEARCH · Search NASA

Results for “Generative artificial intelligence”

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.

652 records · Page 4
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Tokamak Energy’s pre-concept design for a fusion power plant: an overview of ST-E1

Climate change and rapidly rising energy demand, driven in part by artificial intelligence and data-centre growth, create an urgent need for stable, low-carbon, and abundant power. Fusion is a promising long-term solution, yet its commercialisation faces a fundamental paradox in today’s investment environment: pilot plants are essential to de-risk physics, engineering, and operations, but their limited lifetime energy output and high upfront costs make them difficult to finance. This paper presents Tokamak Energy’s response: ST-E1, a pre-concept design for a low-aspect-ratio tokamak power plant engineered specifically to overcome this challenge. ST-E1 is designed from the outset for phased operation—pilot and commercial phases, with an upgrade phase in between—with emphasis on commercial viability, maintainability, nuclear engineering, modularity, and upgradability. A key design principle is the deliberate separation of long-lived assets, such as the magnet cage and vacuum vessel, from replaceable in-vessel systems. This provides an attractive and credible investment approach to generate operational data and de-risk key technologies while preserving most capital-intensive assets for later commercial phases. The architecture supports continuous optimisation toward high net electric power (targeting 800–1000 MW net electric), a normalised capital expenditure of $\$$ 12–14k/kW of net electric power, and high availability (targeting > 80%). A tokamak core with a 5 m major radius, aspect ratio of 2.3, and on-plasma axis toroidal field of 5.25 T was selected to meet these objectives. This paper summarises the ST-E1 design philosophy, principal features, and development methodology. It introduces a Focus Collection of 11 papers detailing the pre-concept design of the entire tokamak and corresponding plant.

ST-E1↗
DOE OSTI2026

Numerical-heating effects in atmospheric pressure streamer discharges simulated with a PIC code

Artificial heating in plasma simulations is a well-known phenomenon which occurs when, among other things, the Debye length is poorly resolved by the simulation mesh. Here, in this work, the degree to which numerical-heating occurs during a simulation of a nanosecond atmospheric pressure streamer discharge is examined. The streamer is simulated using a two-dimensional finite-element, particle-in-cell code Empire, which uses direct simulation Monte Carlo for binary particle interactions. Initially, an estimate of the numerical-heating rate applied to Empire is performed using a simple plasma model. Second, a positive atmospheric pressure streamer discharge simulation is performed to study the effects of numerical heating on plasma density, electron temperature, and streamer velocity. The nominal Debye length is approximately 1 μm and the amount of numerical heating introduced in the simulation is varied by using mesh sizes ranging from 2 μm to 20 μm. A measurable numerical heating quantity is proposed that can be used to estimate the appropriate element size and quantify the numerical-heating that can be expected over the simulation time for an atmospheric pressure streamer. In conclusion while Δx/λ D violations can be an issue it is not likely to be an issue with streamer discharges that are temporally short and occur in environments where collision frequencies are high. This result validates the rationale of grid size choices for a large amount of previously published works where Δx/λ D violation was not clearly addressed. Primary finding of this work is that numerical heating is of minor concern for plasma simulations where electron–neutral collisions are numerous such that multiple collisions can occur within a single plasma period.

Nikic, Dejan [University of New Mexico, Albuquerqu↗
DOE OSTI2026

Roadmap and Benchmarking: Privacy in Federated Load Forecasting

Data-driven techniques for energy demand forecasting continue to emerge with promising impacts on distribution grid planning. However, the development of robust and generalizable machine learning models requires that representative high quality training data are available. Distributed energy resources have begun to embed intelligence, gathering large amounts of data on customer demand, behavior, and household devices that are connected to the grid. Though utilities aggregate meter-level demand data for load shaping, demand response, outage management, reliability planning, and billing applications, there lies an inherent privacy concern in sharing consumption data that may identify individual consumer behavioral patterns. Hence, while sharing the data is crucial, the private sensitive customer data must be safeguarded from being exposed or manipulated. In this study, we propose a roadmap for implementing a based privacy preserving framework to support the advancement of data-driven analytics in data-sensitive distributed energy resources environments. The roadmap incorporates federated learning–a distributed training framework, differential privacy–a statistical framework that provides guarantees to safeguard the leakage of sensitive data, secure multiparty computation and homomorphic encryption– techniques for encrypting model gradients and applying secure aggregation on the server. Moreover, we perform baseline experiments on the federated short-term load forecasting (STLF) task using open-source residential load profile datasets, offering insights into the challenges of integrating differential privacy into federated learning.

Abebe, Waqwoya [Oak Ridge National Laboratory (ORN↗
NASA NTRSDate not supplied

Using Open Innovation in Reducing Risk to Crews

In the exploration of destinations outside of Earth's neighborhood, specifically Mars, scientific and engineering inquiries have occurred by two means; observations from satellites and observations by landed spacecraft. Satellite observations (Mariner, MRO, Mars Odyssey, provide global-scale spatial and temporal data while landed spacecraft (Viking, Mars Pathfinder, Spirit, Opportunity, Phoenix Mars Lander) investigate highly localized areas of the surface of the planet. In preparation for human exploration, extensive knowledge of the surface and atmospheric environments should be known before the first human leaves Earth. The primary goal of performing reconnaissance on Mars on a sub-global scale is to know as much as possible about the environment to which crews will be subjected. At the current rate of launching and landing probes to Mars, it will take a very long time to understand the surface and atmospheric conditions associated with the regions where prospective crews may land. Meanwhile electronics and electrical systems are rapidly getting smaller. One can argue that to acquire the knowledge of the region, one must take hundreds, maybe thousands of measurements simultaneously. One means to perform such a task is to deploy a swarm of sensors. Such a swarm would perform an in-situ assessment of the region. Imagine a close flyby mission to Mars for example, where mini- to micro-sensors are deposited into the atmosphere over half an orbit or more. The sensors, captured by the atmospheric drag and Martian gravity slowly descend buffeted about by Martian winds and weather until they settle on the surface a great time later (think of how long dust takes to settle). As they descend they communicate a vast array of data; temperature, chemistry, pressure, radiation dose, electric or magnetic properties from a region of the planet and an individual sensor need not measure the same quantity as its neighbors. Initially, they could move at the whim of the environment but later versions could have locomotion or propulsion mechanisms. Humans wouldn't need to decide where the sensors go, the sensors do that for themselves. This is a key strength of a sensor swarm. The intelligence relies on the group not on a decision maker on earth. Real time sensor inputs direct what the swarm considers most interesting to investigate resulting in emergent behavior. We issued a $20,000 challenge to the global innovators to provide solutions as to how such a swarm could be initialized and by what protocols and methodologies by which they operate. Over 400 innovators from 49 countries took a look at the problem, with three receiving partial awards for solutions.

Mel Ferebee↗
DOE OSTI1968

Nuclear Physics Made Very, Very Easy

The fundamental approach to nuclear physics was prepared to introduce basic reactor principles to various groups of non-nuclear technical personnel associated with NERVA Test Operations. NERVA Test Operations functions as the field test group for the Nuclear Rocket Engine Program. Nuclear Engine for Rocket Vehicle Application (NERVA) program is the combined efforts of Aerojet-General Corporation as prime contractor, and Westinghouse Astronuclear Laboratory as the major subcontractor, for the assembly and testing of nuclear rocket engines. Development of the NERVA Program is under the direction of the Space Nuclear Propulsion Office, a joint agency of the U. S. Atomic Energy Commission and the National Aeronautics and Space Administration. This report is being reprinted for use in the U. S. Atomic Energy Commission and National Aeronautics and Space Administration educational and technology utilization programs.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
NASA NTRS1993

Performance Results of a Digital Test Signal Generator

Performance results of a digital test signal-generator hardware-demonstration unit are reported. Capabilities available include baseband and intermediate frequency (IF) spectrum generation, for which test results are provided. Repeatability in the setting of a given signal-to-noise ratio (SNR) when a baseband or an IF spectrum is being generated ranges from 0.01 dB at high SNR's or high data rates to 0.3 dB at low data rates or low SNR's. Baseband symbol SNR and carrier SNR (P c /N o ) accuracies of 0.1 dB were verified with the built-in statistics circuitry. At low SNR's that accuracy remains to be fully verified. These results were confirmed with measurements from a demodulator synchronizer assembly for the baseband spectrum generation, and with a digital receiver (Pioneer 10 receiver) for the IF spectrum generation.

B O Gutierrez-Luaces↗
DOE OSTI2026

A generative machine learning model for designing metal hydrides applied to hydrogen storage

Developing new metal hydrides is a critical step toward efficient hydrogen storage in carbon-neutral energy systems. However, existing materials databases, such as the Materials Project, contain a limited number of well-characterized hydrides, which constrains the discovery of optimal candidates. This work presents a framework that integrates causal discovery with a lightweight generative machine learning model to generate novel metal hydride candidates that may not exist in current databases. Using a dataset of 450 samples (270 training, 90 validation, and 90 testing), the model generates 1000 candidates. After ranking and filtering, six previously unreported chemical formulas and crystal structures are identified, four of which are validated by density functional theory simulations and show strong potential for future experimental investigation. Overall, the proposed framework provides a scalable and time-efficient approach for expanding hydrogen storage datasets and accelerating materials discovery.

generative model↗
NASA NTRS1965

Computation of Optimal Interplanetary Low-Thrust Trajectories With Bounded Thrust Magnitude By Means of the Generalized Newton-Raphson Method

The generalized Newton-Raphson method, an iterative procedure for solving nonlinear operator equations, has been extended in application to variational problems with bounded control variables. A minimum fuel interplanetary low thrust orbital transfer problem is worked out in detail to demonstrate the practical aspects of the algorithm as well as its computational effectiveness. The control variables are the thrust magnitude, limited from zero to some prescribed maximum value, and the thrust steering angle.

Computation↗
DOE OSTI2026

From Rules to Reasoning: A Survey of Large Language Model-Based Approaches to Scientific Hypothesis and Idea Generation

Scientific hypothesis generation represents a fundamental challenge in contemporary research due to exponentially expanding literature volumes and increasing disciplinary specialization. Large language models (LLMs) have emerged as transformative tools for automated scientific discovery, moving beyond traditional rule-based and literature-mining approaches. Four paradigmatic approaches define current LLM-driven hypothesis generation: direct prompting and fine-tuning methods, knowledge-enhanced frameworks integrating retrieval-augmented generation (RAG), multi-agent collaborative systems simulating research teams, and reasoning-focused approaches implementing cognitive architectures. Domain-specific applications demonstrate statistical equivalence to human expert performance in social psychology, experimental validation in biomedical research, and near-expert quality in astronomy. Evaluation methodologies encompass human expert assessment, LLM-as-judge frameworks, and comprehensive benchmarking systems. Technical challenges include hallucination management, knowledge integration limitations, and balancing novelty with feasibility. Future directions emphasize hybrid neural-symbolic architectures and sophisticated human-AI collaboration models for responsible scientific discovery acceleration.

AI-driven discovery↗