Search NASA⌕ Search

SEARCH · Search NASA

Results for “performance objectives”

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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Transforming jet flavour tagging at ATLAS

Jet flavour tagging enables the identification of jets originating from heavy-flavour quarks in proton–proton collisions at the Large Hadron Collider, playing a critical role in its physics programmes. This paper presents GN2, a transformer-based flavour tagging algorithm deployed by the ATLAS Collaboration that represents a different methodology compared to previous approaches. Designed to classify jets based on the flavour of their constituent particles, GN2 processes low-level tracking information in an end-to-end architecture and incorporates physics-informed auxiliary training objectives to enhance both interpretability and performance. Its performance is validated in both simulation and collision data. The measured c-jet (light-jet) rejection in data is improved by a factor of 3.5 (1.8) for a 70% b-jet tagging efficiency, compared to the previous algorithm. GN2 provides substantial benefits for physics analyses involving heavy-flavour jets, such as measurements of Higgs boson pair production and the couplings of bottom and charm quarks to the Higgs boson, and demonstrates the impact of advanced machine learning methods in experimental particle physics.

Characterization and analytical techniques↗
DOE OSTI2025

Analyzing and Exploring Training Recipes for Large-Scale Transformer-Based Weather Prediction

Abstract The rapid rise of deep learning (DL) in numerical weather prediction (NWP) has led to a proliferation of models which forecast atmospheric variables with comparable or superior skill than traditional physics-based NWP. However, among these leading DL models, there is a wide variance in both the training settings and architecture used. Further, the lack of thorough ablation studies makes it hard to discern which components are most critical to success. In this work, we show that it is possible to attain high forecast skill even with relatively off-the-shelf architectures, simple training procedures, and moderate compute budgets. Specifically, we train a minimally modified Swin Transformer V2 (SwinV2) on ERA5 data and find that it attains superior skill in terms of mean-square errors of deterministic forecasts when compared against the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS). Almost all DL–NWP systems share a core set of hyperparameters and design decisions. To aid and expedite future DL–NWP research, we present an in-depth, systematic exploration of different loss functions, model sizes and depths, patch sizes, and multistep training objectives. We also examine the model performance with metrics beyond the typical accuracy (ACC) and RMSE and investigate how the performance scales with model size. Through our open-source code, scoring pipelines, and models, we share our findings on key aspects of the training pipeline. These ablations reduce the necessity for expensive hyperparameter tuning and lower the barrier to entry for future DL–NWP research. Significance Statement This study investigates the potential of using large-scale transformer-based models for weather prediction, showing that it is possible to achieve high forecast accuracy with simpler, off-the-shelf architectures. By training a minimally modified SwinV2 transformer on ERA5 data, we show that the model achieves competitive forecast skill in terms of mean-square error for key variables, outperforming the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS) at all lead times. Our findings suggest that effective training strategies, such as multistep fine-tuning and channel-weighted losses, significantly enhance the model’s performance. However, we also highlight that these improvements come with trade-offs in other areas, such as ensemble spread and high-frequency spatial detail. This work highlights the promise of deep learning in improving weather forecasts, which could lead to better preparedness and response to weather events, ultimately benefiting society by providing more reliable weather predictions.

Willard, Jared D. [Lawrence Berkeley National Labo↗
DOE OSTI2024

Testing- and Model- Based Optimization of Coal-fired Primary Heater Design for Indirect Supercritical CO 2 Power Cycles (Final Scientific and Technical Report)

The overall objective of this project was to perform the R&D necessary to mitigate the risk associated with the design of a primary heat exchanger for a solid-fired combustion system coupled with an indirect-fired closed-loop Brayton Cycle utilizing supercritical CO 2 . The key technological hurdle was the coupling of a solid-fuel firing system with the primary heater, which poses a singular challenge, which is the management of burner performance and operational conditions in a way to manage heat exchanger tube metal temperatures and temperature ramp rates in the absence of fluid phase change on the inside of the tubes. We designed and built the first ever pseudo power system employing a simple recuperated supercritical CO 2 closed-loop Brayton Cycle coupled to a solid-fuel fired system. Advanced coupled CFD and process modeling were used to design the primary heat exchanger (PHX), which consisted of both radiative and convective sections, to limit tube metal temperatures resulting from the heat release profile of the solid fuel flame near the radiative tubes. The heat exchanger was designed to produce finished CO 2 temperatures of 600 °C a pressure of 20.7 MPa and CO 2 flow of 5.5 kg/s. The constructed PHX was capable of 1.2 MWth heat uptake. During design of the PHX, the modeling showed that most variables influencing flame shape (burner stoichiometric ratio and register velocities and swirl) were not suitable to manage heat flux to the metal surfaces. This is because they substantially increased adiabatic flame temperature through the influence of localized stoichiometric ratio. Excess air and firing rate were the two most powerful variables that could be used to control tube surface temperatures. The coupled system was operated for a total of 407 hours, with the longest continuous run of 248 hours. For 62% of the operational time, the unit was unmanned and in automatic control. The fuels used for the testing included natural gas, two Utah Bituminous coals, woody biomass, and bagasse. During the testing we were able to verify the 1.2 MWth heat uptake and we operated at a finished CO 2 temperature of 607 °C and a pressure of 20.3 MPa simultaneously. The real-time corrosion rate of the Super 304H tube CO 2 surface in the region of the radiative section of the PHX were measured, at an approximate temperature of 550 °C. The two key variables related to corrosion rate are the pressure and flow rate of the CO 2 . A technoeconomic analysis was performed at a scale of 120 MWE. The updated analysis showed that the efficiency of an sCO 2 power producing plant will be related to the pressure drop of the PHX.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2026

Brochure for the DOE Office of Science Workshop on Envisioning Frontiers in AI and Computing for Biological Research

In February of 2025 a joint ASCR/BER workshop was held to identify key transformational research directions for understanding biology using artificial intelligence (AI), digital twins and high-performance (HPC) computational methods to facilitate scientific discovery and innovation in support of the Department of Energy mission. AI technologies offer exciting new groundbreaking methods to analyze large volumes of complex biological data, thereby greatly accelerating the ability to understand, predict, and design biological processes for beneficial purposes. In the laboratory, the bridging of AI-enabled automated experimental technologies, HPC and digital twins will provide potent tools for researchers to explore the fundamental nature of biology and harness its inherent metabolic potential for a variety of beneficial purposes. The focus of this workshop was on how high-performance computational methods can impact this objective by exploring digital twins, foundational models, and data-driven approaches with applications to advance automated laboratory experiments, modeling of complex living systems and engineering new functions into plants and microbial systems relevant to DOE mission. Workshop attendees with expertise in plant science, microbiology, mathematics, computer science, and AI assessed the current state of the science, trends, and AI challenges at the interface of plant and microbial systems biology and computational science to identify opportunities for high-impact research. This collaborative effort capitalized on ASCR's advancements in applied mathematics, computer science, and Exascale systems, and BER's expertise in basic genomics-enabled research on DOE relevant plant and microbial systems. The workshop culminated in four key priority research directions to guide future research and development within DOE Office of Science programs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Non-catalytic pyrolysis of associated gas to zero CO 2 hydrogen and high value carbon black

The overall objective of this work is to perform a pre-Front End Engineering Design (pre-FEED) study of the conversion of associated gas produced at an active oil pad of a Bakken oil field producer, to high value carbon black and hydrogen using the Microwave Plasma Pyrolysis of Associated Gas (MPP-AG) process developed by H Quest Vanguard, Inc (HQV). The University of North Dakota’s Center for Process Engineering Research (CPER) collaborated with HQV to complete a design and techno-economic analysis for the MPP-AG deployed at an active Bakken well site flaring an average of 54,000 standard cubic feet per day of AG. The proposed design consists of: • A Microwave Pyrolysis Unit (MPU) which consists of H Quest’s proprietary MPP technology • A patent-pending Gas Conditioning Unit (GCU) designed by the University of North Dakota to condition intermittent and variable flow gas prior to pyrolysis. • An auxiliary unit running on the associated gas to power the GCU and MPU in an “island mode” configuration.

03 NATURAL GAS↗
DOE OSTI2025

Boiling Water Capsule Irradiation Test Device Design Study Report

RELAP5 calculations were conducted to support an experimental design, referred to as water capsule, which can facilitate fuel performance testing within ATR. The main objective is to provide useful insights into the experimental design and operating strategies that can mimic LWR operating conditions within the spatial constraints of the ATR position under consideration. This report summarizes the RELAP5 analysis results based on the most recent water capsule design derived through various tests. To discuss a design under conservative conditions, fuel power is consistently assumed to be 400 W/m in all analyses.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2026

Final Technical Report

The overall objective of the project was to perform the preliminary development work for an integrated CO2 capture project at the Polk Power Station. During the project, the team applied its extensive knowledge base, built from previous Front End Engineering and Design (FEED) studies and recent power generation major construction projects, to complete the conceptual design of a carbon capture system capable of capturing a minimum of 95% of the CO2 emissions.

Dilport, Nathan↗
DOE OSTI2024

Isotope Production Education and Research via a Systematic Study of Photo-nuclear Reaction Yields and Excitation Functions

A new collaboration between the New Mexico Institute of Mining and Technology, Idaho State University, and Idaho National Laboratory builds the domestic workforce in isotope production and related science and technologies by conducting research and education programs related to photo-nuclear reactions. Workforce development was identified by the Department of Energy’s Office of Science in the Reaching a New Energy Sciences Workforce (RENEW) program that seeks to engage underrepresented university students in isotope production activities. This collaboration includes coursework development, student training, and isotope production research utilizing photo-nuclear reactions at electron linear accelerators. Bremsstrahlung-weighted excitation functions and cross-sections will be measured and the use of nanomaterials to exploit kinematic recoil reactions will be investigated. Students will be hosted at the Idaho National Laboratory where mentors will provide opportunities to explore careers in nuclear science and technology. This presentation regards the ongoing progress of the research and education objectives of the collaboration that was performed under the Office of Science Isotopes Program under award number D000-22-2765.

07 ISOTOPE AND RADIATION SOURCES↗
DOE OSTI2024

Economic and Jobs Impacts of Point-Source Carbon Capture in Cement Industry – Case Study

The cement industry accounts for an estimated 8% of global CO2 emissions, which surpasses that of the entire aviation sector. In contrast with other industries, where CO2 emissions can be drastically reduced via electrification or fuels substitution, cement production releases CO2 as part of its process, during the calcination of carbonates to yield oxides. Thus, point-source carbon capture has become a key technology in the cement industry’s decarbonization. Apart from the expected environmental benefits, point-source carbon capture in the cement industry can yield important economic benefits and create jobs. The objective of this study was to perform a preliminary assessment of the economic and workforce impacts associated with the construction and operation of a point-source carbon capture retrofit of an existing cement production facility, using as a basis the data from a front-end engineering design (FEED) study to install a 3.9 million metric tons per year (Mtpy) CO2 capture facility at Holcim Ste Genevieve cement plant in Missouri, United States of America. The advanced carbon capture technology used in this FEED study was Air Liquide’s Cryocap™ FG carbon capture technology. The study evaluated the direct, indirect, and induced economic impacts of the construction, operation, and maintenance activities of the project over its lifespan. It also covered how the project will generate new jobs, their nature, and quantity, along with strategies to prepare the workforce. To perform this study, construction, operation, and maintenance cost estimates, as well as construction and operation staffing plans from the FEED study were input into IMPLAN version 7.5 software, licensed by IMPLAN Group LLC (Huntersville, VC), to predict the direct, indirect and induced economic impacts of the project using industry multipliers from the software. Additionally, recruitment strategies were developed for hiring individuals who belong to groups that are historically underserved or underrepresented, as well as anticipated recruitment of workers from the local community (whether training will be required or if the skills are associated with an existing labor force). The analysis estimated that the construction and operation of the carbon capture at Holcim Ste. Genevive will result in over 24 thousand work-years of job opportunities, close to USD 10 billion of economic impacts, including over USD 460 million of tax revenue. These results encompass the direct, indirect, and induced effects. A strategy to maximize hiring from the project and neighboring counties was developed, leveraging training agreements with local trade groups and universities. The result of this study can be used for a strategic preliminary assessment of the potential regional economic and job impacts of retrofitting existing cement plants with point source carbon systems, and its methodology can be replicated to individual projects to aid in planning and workforce development.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2024

Experimental investigation of thermal conductivity of high entropy ceramics

In this talk, we will present the experimental investigation of thermal conductivity of high entropy carbides (HEC) and borides (HEB) with different compositions. Thermal conductivity determines the capability of a material to transport thermal energy and is considered as one critical physical property, specially in the energy industry. Due to the presences of disordered cation sublattice, it is expected that thermal conductivity of high entropy ceramics is lower than the counterparts with simpler compositions, but the experimental studies are relatively few. Using laser-based spatial-domain and frequency-domain thermoreflectance technique, we measured thermal conductivity of HECs and HEBs in a temperature range from 77K to 773K. Furthermore, by focusing the laser using an objective lens, we were able to perform the thermal property mapping on samples with the spatial resolution of micrometer. Such measurements could reveal important microstructure information in a rapid and efficient way without using scanning electron microscopy. The high spatial resolution also enables thermal conductivity measurements on ion-irradiated HECs with the damage layer of merely 1.1µm. These measurements can provide valuable information to predict the performances of HECs and HEBs with different compositions and from different manufacturing processes, and in the extreme environments such as nuclear reactors.

36 - MATERIALS SCIENCE↗
DOE OSTI2026

ECAR-8848 Rev 2 MARVEL Primary Coolant System Lifting Analysis

The objective of this ECAR is to perform a lifting analysis of the Microreactor Applications Research Validation and Evaluation (MARVEL) Primary Coolant System (PCS) vertically, pickup from horizontal, the temporary lifting plate, and spreader.

21 - SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLAN↗
DOE OSTI2024

Concept Study of Robotic Camera-Based Foreign Object Detection for EV Wireless Charging

Wireless charging of an electric vehicle (EV) is an emerging charging technology promising convenient, autonomous, and highly efficient EV charging without requiring heavy gauge cables. However, due to the strong electromagnetic field created by this process that surrounds the wireless charger, the presence of foreign objects can detrimentally interact with it, thus affecting wireless power transfer (WPT) performance or leading to harmful and unwanted safety risks. This paper presents the results for a concept study on a robotic camera-based foreign object detection (FOD) system, as a supplement to the industry-existing overlapped FOD coil array method, for EV wireless charging. A Raspberry PI 4 control board and compatible Raspberry PI Camera Module 2 are used to implement camera-based object detection. The FOD program was developed using a state-of-the-art deep learning object detection model with the OpenCV and Pytorch library and is compatible with camera module hardware. A dry-run test with Raspberry PI and a camera module was conducted and the preliminary FOD function was verified. The feasibility assessment is also validated by comparing the performance of five existing state-of-the-art deep learning object detection models for vehicles, animals, persons, and metals subsets, respectively. Satisfactory performance on the benchmark datasets is observed by the tests, but further improvements are needed in future work when detecting small-sized metallic objects. A programable robotic car is also under development as ongoing work for carrying the Raspberry PI and camera module while moving for the maintenance process.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

Optimization of R290 variable geometry heat exchangers

Air-to-refrigerant heat exchangers (HX) are vital components in Heating, Ventilating and Air Conditioning and Refrigeration (HVAC&R) equipment. Recent literature has proposed utilizing shape-optimized, non-round tubes to reduce component size and refrigerant charge, thus enabling the adoption of low-GWP natural refrigerants like R290. However, most designs are restricted to fixed tube configurations, limiting the overall performance potential. In this work, multi-objective optimization is used in a staged approach to develop variable geometry multi-pass air-to-R290 HXs with minimal volume and airside pressure drop. Non-round tubes are used throughout the domain and each pass may have different tube arrangements to maximize HX performance. Compared to conventional fin-tube HXs, the optimized designs demonstrate more than 60% reductions in envelope volume, 30% reductions in face area, and up to 24% reduction in airside pressure drop. Additionally, the refrigerant charge reduction was up to 64%, thus enabling safe use of flammable and mildly-flammable refrigerants.

42 ENGINEERING↗
DOE OSTI2025

Relationships Between Mesoscale Convective System Properties and Midlevel Dynamic Perturbations

Abstract Past studies implicate dynamic anomalies operating on subsynoptic scales as a possible initiation source of summertime (July–August) mesoscale convective systems (MCSs) in the central United States during northwesterly flow regimes. To improve our understanding of warm season MCSs occurring over a variety of flow regimes, we track midlevel (600 hPa) vorticity perturbations (“MPs”) as 2D objects comprising wavelengths of 500–2,500 km over the central US from May–August of 2004–2021. We perform statistical analysis of relationships between metrics of MP objects (e.g., duration, size, intensity, and origin) and high‐resolution MCS precipitation characteristics (e.g., duration, total rainfall, rain coverage area, and motion) that occur while collocated with or in the absence of MPs to discern predictive capability of background dynamic features on storm precipitation potential. Although the majority of MPs collocated with MCS initiation occur during July–August, a significant number (40%) occur between May and June. Northwesterly flow MPs comprise a relative minority of our events, suggesting that MPs can affect MCSs across a variety of warm season flow regimes. MPs affecting MCSs initiated primarily over the high plains near the central Rockies. Only approximately 20% of tracked MCS initiation events were collocated with MPs, but these storms produced ∼25% greater lifetime rainfall and coverage area, and ∼29% more stratiform rain than non‐MP‐induced MCSs. In general, larger and more vigorous MPs resulted in more hydrologically impactful MCSs. The most directly attributable benefit to MCS initiation was from MP‐enhanced background vertical motion and thermodynamic instability (e.g., increased CAPE).

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Distributed Stochastic Optimization of a Neural Representation Network for Time-Space Tomography Reconstruction

4D time-space reconstruction of dynamic events or deforming objects using X-ray computed tomography (CT) is an important inverse problem in non-destructive evaluation. Conventional back-projection based reconstruction methods assume that the object remains static for the duration of several tens or hundreds of X-ray projection measurement images (reconstruction of consecutive limited-angle CT scans). However, this is an unrealistic assumption for many in-situ experiments that causes spurious artifacts and inaccurate morphological reconstructions of the object. To solve this problem, we propose to perform a 4D time-space reconstruction using a distributed implicit neural representation (DINR) network that is trained using a novel distributed stochastic training algorithm. Our DINR network learns to reconstruct the object at its output by iterative optimization of its network parameters such that the measured projection images best match the output of the CT forward measurement model. Here, we use a forward measurement model that is a function of the DINR outputs at a sparsely sampled set of continuous valued 4D object coordinates. Unlike previous neural representation architectures that forward and back propagate through dense voxel grids that sample the object's entire time-space coordinates, we only propagate through the DINR at a small subset of object coordinates in each iteration resulting in an order-of-magnitude reduction in memory and compute for training. DINR leverages distributed computation across several compute nodes and GPUs to produce high-fidelity 4D time-space reconstructions. We use both simulated parallel-beam and experimental cone-beam X-ray CT datasets to demonstrate the superior performance of our approach.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Enabling the Broader Use of MOOSE for Nuclear Energy and Other Simulation

This Final Scientific and Technical Report summarizes work performed under the Phase IIA SBIR project “Enabling the Broader Use of MOOSE for Nuclear Energy and Other Simulation” (DE-SC0020906) from August 2023 through August 2025. The objective of the Phase IIA effort was to mature and harden capabilities developed during Phase II, with the goal of enabling practical interoperability between Coreform’s isogeometric analysis (IGA) technologies and the Multiphysics Object-Oriented Simulation Environment (MOOSE), while improving robustness, performance, and scalability for complex, nuclear-relevant geometries. Over the course of Phase IIA, the project established and validated an extraction-based interoperability pathway between Coreform tools and MOOSE. A combined mesh and matrix format was defined collaboratively with MOOSE developers and integrated into the solver, enabling standard MOOSE workflows to operate on data exported from Coreform’s IGA and Flex Representation Method (FRM) pipelines. Early demonstrations validated architectural compatibility using linear solid mechanics problems, while later efforts focused on benchmark testing and external use. By the end of the project period, engineers at BWXT were able to independently set up and execute a simulation using the Coreform–MOOSE workflow and provide direct feedback that informed further refinement. In parallel, substantial effort was devoted to improving the robustness of trimmed U-spline construction for complex CAD geometries. A growing test suite of nuclear-relevant models was compiled through collaboration with multiple stakeholders and used to drive extensive bug fixing and reliability improvements. These efforts resulted in improved robustness and performance, including the addition of fallback capabilities that enhance reliability when the underlying commercial CAD kernel fails. Performance-oriented work progressed later in the project, with the development and demonstration of methods to decompose complex geometries into structured subregions and updated data representations to support more efficient solver processing. Additionally, extensive enhancements to threadsafe parallel data structures and trimming operations established a foundation for scalable processing of large assemblies. Collaboration with Sandia National Laboratories on the SGM geometric modeling kernel advanced to a functioning interface test case, positioning the workflow for future kernel integration. Overall, the Phase IIA effort successfully transitioned the project from architectural proof-of-concept to externally exercised, solver-integrated capability, while clarifying remaining technical challenges related to standardization, performance optimization, and kernel integration.

42 ENGINEERING↗
DOE OSTI2025

AGR-5/6/7 Irradiation Test Final As-run Report

This document presents the as-run analysis of the Advanced Gas Reactor (AGR)-5/6/7 irradiation experiment. AGR-5/6/7 is the last of a series of experiments conducted in the Advanced Test Reactor (ATR) at Idaho National Laboratory in support of the development and qualification of tri-structural isotropic low-enriched fuel for use in high-temperature gas-cooled reactors. The test train contained five separate capsules that were independently controlled and monitored. Each capsule contained multiple 24.91-mm-long and 12.25-mm-dimeter compacts filled with low-enriched uranium carbide/oxide tri-structural isotropic fuel particles. The objectives of the AGR-5/6/7 experiment were to: • Irradiate reference-design fuel particles to support fuel qualification. • Establish operating margins for the fuel, beyond normal operating conditions. • Provide irradiated-fuel performance data and irradiated-fuel samples for post-irradiation examination and safety testing. The primary objective of the AGR-5/6 test (Capsules 1, 2, 4, and 5) was to verify the successful performance of the reference-design fuel under normal operating conditions. The AGR-7 test (Capsule 3) was designed to explore fuel performance at higher temperatures. Its primary objective was to demonstrate the capability of the fuel to withstand conditions beyond normal operating conditions, in support of plant design and licensing. AGR-5/6/7 will also provide irradiated-fuel performance data based on the fission gas release from particles during irradiation. To achieve the test objectives, the AGR-5/6/7 experiment was irradiated in the northeast flux trap of the ATR with a planned duration of 500 effective full-power days. The northeast flux trap was selected because its larger diameter provided greater flexibility for test-train design compared to the Large B positions used for the AGR-1 and AGR-2 irradiations, significantly enhancing test capabilities for the combined irradiation campaigns. Due to delays in the ATR schedule, the AGR-5/6/7 irradiation was significantly shorter than the originally planned 13-cycle schedule. Irradiation began on February 16, 2018 and ended on July 22, 2020, spanning nine ATR cycles (162B–168A) over two and a half years. Thus, the AGR-5/6/7 fuel compacts were irradiated for a total of approximately 360.9 effective full-power days. Final burnup values, on a per-compact basis, ranged from 5.66 to 15.26% fissions per initial heavy metal atom, while fast fluence values ranged from 1.62 to 5.55 × 1025 n/m2 (E >0.18 MeV). Time-averaged volume-averaged fuel temperatures on a capsule basis at the end of irradiation ranged from 756°C in Capsule 5 to 1313°C in Capsule 3 excluding days with significantly lower temperature during the two short powered axial locator mechanism cycles, 163A and 167A. By the end of irradiation, 48 out of 54 installed thermocouples had failed (the bottom three capsules lost all thermocouples). During the first five cycles (162B – 165A), the fission-gas isotope release-rate-to-birthrate (R/B) ratios were stable in the 10-8–10-6 range, and no in-pile particle failures were observed based on the gross gamma counts. During this time, the high exposed kernel fraction and high fuel particle temperatures in Capsule 1 led to a maximum R/B value of around 2 ? 10-6 for Kr-85m. The fission gas release in all capsules started to increase from the second half of Cycle 166A, when a large number of in-pile particle failures occurred in Capsule 1 and a gas line problem in this capsule caused fission gas leakage at various degrees into the other four capsules. This gas line problem also prevented a fission gas release measurement for Capsule 1 during the last three cycles due to gas flow isolation. By the end of irradiation, it is estimated that approximately 15 particles failed in Capsule 3, which was considered possible because the experiment was designed to operate beyond the high-temperature gas-cooled reactor normal operating temperature range. A few hundred in-pile particle failures were estimated for Capsule 1 by the end of Cycle 166A, but the total number of failures is unknown due to the lack of fission gas release data in the later cycles. Additionally, four potential in-pile failures were identified for Capsule 2 during the last cycle, Cycle 168A. In contrast, no in-pile failures were identified in the top two capsules (4 and 5) based on the absence of the typical spikes in gross gamma counts and low failure estimates using the AGR-3/4 R/B per exposed kernel model.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2026

HARMONY: Large-Scale Architecture Search for Efficient Hybrid Language Models

As large language models scale to trillions of parameters, their computational and memory requirements present critical challenges for efficient training and deployment. While Mixture of Experts (MoE) architectures enable efficient scaling through sparse parameter activation, and state-space models like Mamba offer linear-time complexity, principled methods for combining these paradigms remain undeveloped. We introduce HARMONY (Hybrid Architecture Research for Mamba, Optimized with Neural efficiencY), a multi-objective evolutionary neural architecture search framework for discovering efficient hybrid language models that integrate Transformer attention mechanisms, Mixture-of-Experts routing, and Mamba state-space components. Through large-scale distributed search using 16,384 MI250X GPUs on the Frontier supercomputer, HARMONY explores a comprehensive design space encompassing six attention variants (MHA, MQA, GQA, MLA, SWA, and Mamba-2), variable MoE configurations with both routed and shared experts, and extensive Mamba hyperparameters. Our framework discovers heterogeneous architectures that balance training performance with computational efficiency through multi-objective optimization incorporating latency penalties and fitness-based selection. Analysis of discovered architectures reveals that optimal hybrid designs favor heterogeneous component mixing rather than homogeneous patterns, with Mamba-2 and Multi-Head Latent Attention (MLA) emerging as preferred mechanisms. Discovered architectures demonstrate superior training efficiency: our best configuration achieves a final perplexity of 1.0874 with 2.38B parameters while processing 4,320 tokens/second, outperforming significantly larger manually designed models. Full-scale evaluation shows HARMONY's top architectures achieve better loss trajectories than equivalently-sized models using state-of-the-art configurations including Mixtral, Jamba, and Samba. Additionally, we demonstrate 91% weak scaling efficiency when training discovered 36B-parameter models across 1,024 GPUs. HARMONY is released as an open framework with comprehensive tools for building and training hybrid models using expert-data-pipeline parallelism, democratizing access to automated architecture design for next-generation language models.

Herron, Emily [ORNL] (ORCID:0000000273008172)↗