Search NASA⌕ Search

SEARCH · Search NASA

Results for “adaptive machine learning”

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 379 records · Page 21
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Project Development of an Electrochemical Denitration and Caustic Generation System for HLW Pretreatment at Hanford - 26350

An engineering-scale electrochemical processing skid is proposed to perform the denitration of Hanford tank waste, which would help to mitigate a key process concern with the direct feed processing of the Hanford Tank Waste Treatment and Immobilization Plant (WTP). The reduction of nitrates and organic compounds in the waste feed will directly reduce hazardous NOx and ammonia gases generated during the vitrification process, which in turn will aid in addressing potential regulatory and safety challenges associated with processing large volumes of tank waste. This paper highlights the past legacy work, project layout, accomplishments from Phase 1 and research and development envisioned for Phase 2. An innovative electrochemical denitration and caustic generation (EDCGe) process was demonstrated for the pretreatment of tank waste at the Savannah River Site (SRS) in the early 2000s. The denitration electrolyzer, off-gas abatement system, and caustic generator electrolyzer are being developed with the intent that the denitration electrolyzer will convert nitrate and nitrite anions to nitrogen gas while also yielding other gaseous byproducts, which may include N2O, NH3, VOCs, and H2. The gaseous byproducts will be managed via a tandem off-gas catalyst-bed treatment system. The caustic generation electrolyzer will recycle NaOH from the feed to produce a clean caustic stream for use within the batching tanks at Hanford, aiding in the preparation of waste for WTP. The reduction in hazardous emissions and improved waste treatment processes provides a robust solution for nuclear waste management, contributing to environmental safety and regulatory compliance. The EDCGe technology is being adapted, modified, and updated for the preparation of the Direct Feed-High Level Waste (DF-HLW) flowsheet at Hanford. Phase 1 demonstrated a bench-scale proof-of-concept for reactions involving the denitration electrolyzer and gas phase abatement of ammonia. The electrochemical technology is drawing on the scientific outcomes that were reported in the legacy work. The results from Phase 1 demonstrated the viability of the EDCGe system in reducing the nitrogen species of simple non-radioactive waste simulants. Commercially available alloys used as electrode materials and membranes are being studied for the denitration and caustic generation electrolyzers. The continuation of this project holds promise for broader applications, such as energy-efficient ammonia production, and contributes significant advancements in nuclear waste management. Additional material discovery has been investigated into ceramic Na super ion conductive (NaSICON) materials and off-gas abatement catalyst discovery. NaSICON is of interest for selective transport of Na within the electrolyzers to make a clean caustic stream. Future integration and optimization efforts, informed by Phase 1 results and ongoing research, will continue to drive advancements in nuclear waste management technology. The technology developed for the EDCGe treatment of tank waste will also have broader potential to inform other fields, such as energy-efficient ammonia production, as well as ammonia abatement catalysis through the lessons learned in electrochemical nitrate reduction. The applications and benefits of this research extend beyond Hanford and the Savannah River Site, supported by a collaborative team of scientists and engineers from national labs, academia, and industry, ensuring a comprehensive approach to solving complex waste treatment challenges. The team is leveraging advanced electrochemical technologies, machine learning, novel catalysts tailored for gaseous nitrogen species, and cutting-edge reactor systems to enhance the process efficiency and effectiveness of the denitration process.

Rodene, Dylan [Savannah River National Laboratory ↗
DOE OSTI2025

Model Residuals as Shields: A Two-Level Formulation to Defend Smart Grids From Poisoning Attacks

The advancement of smart grids presents both vast opportunities and heightened cybersecurity risks. Data-driven defense mechanisms, though designed as a shield against these threats, can fall prey to poisoning attacks. We delve into regression settings, underscoring the imperative to fortify defenses against a spectrum of poison ratios, notably those above 0.5—an issue scarcely addressed in prior studies. Recognizing the susceptibilities of smart grids and their manipulable sensors, we exploit the very intent of poisoning attacks, compromising model accuracy, as our defense mechanism. Our proposed two-level optimization framework discerns between poisoned and authentic data based on model residuals, outperforming or matching existing methods in 72% to 77% of precision and 75% to 80% of recalls across various poisoning attacks, poison ratios, and datasets. Once the authentic data are identified, the trained model is adaptable for a variety of applications. Comprehensive evaluations on different smart grid datasets, pitted against myriad poisoning schemes, validate our methodology’s edge over existing methods. Here, we also shed light on the implications of model misspecification originating from temporal auto-correlation, a common feature in Internet of Things and smart grid data.

Adversarial machine learning (ML)↗
DOE OSTI2024

Combined Machine Learning and Molecular Dynamics Reveal Two States of Hydration of a Single Functional Group of Cationic Polymeric Brushes

The state of hydration of a macromolecular system regulates a plethora of different properties of such a system. In this article, we develop a novel machine learning (ML) approach, based on the unsupervised clustering algorithm, for probing the hydration behavior of the {N(CH 3 ) 3 } + functional group of the PMETAC [Poly(2-(methacryloyloxy)ethyl trimethylammonium chloride] polyelectrolyte (PE) brush system. The PE brushes and the brush-supported water molecules and counterions (chloride ions) are first described using all-atom molecular dynamics (MD) simulations. The simulation data is subsequently used in our ML framework to identify that (1) the {N(CH 3 ) 3 } + functional groups of the PMETAC brushes have two distinct hydration states with one state (state 1) being characterized by less structured water molecules and the other state (state 2) being characterized by more structured water molecules and (2) an enhancement in the brush grafting density leads to the progressive dissapparenace of state 2. An increase in the grafting density increases the number of chloride counterions in a given volume around the {N(CH 3 ) 3 } + functional group and increases the number of shared water molecules between the {N(CH 3 ) 3 } + and Cl - . The chloride counterions are associated with a hydration layer with much less structured water molecules. Therefore, with an increase in the grafting density, an increase in the percentage of shared water molecules leads to the prevalence of the hydration state [of the {N(CH 3 ) 3 } + moiety] with less structured water molecules. Finally, we explain how the present findings are commensurate with two key previous related results, namely a significantly large chloride ion mobility inside the PMETAC brush layer and the {N(CH 3 ) 3 } + -Cl - average distance remaining independent of the PMETAC brush grafting density. Furthermore, we anticipate that the combined ML-MD-simulation approach proposed in this study can be adapted to probe other soft matter systems to reveal new insights of the underlying mechanisms of emergent phenomenon.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Machine learning and process-based modeling of spatiotemporal changes in active layer thickness across Alaska

Permafrost degradation poses a growing threat to infrastructure stability and ecosystem resilience in the rapidly warming Arctic. We investigated the spatiotemporal dynamics of active layer thickness (ALT) across Alaska by integrating field observations, environmental datasets, a physically based Stefan model, and machine learning (ML) techniques. Using weather projections from the Coupled Model Intercomparison Project Phase 6 under two Shared Socioeconomic Pathways (SSP 2-4.5 and SSP 5-8.5), we assessed ALT sensitivity to projected future weather conditions. The random forest (RF) model outperformed the Stefan approach in predicting ALT on the training dataset (R² = 0.84 vs. 0.53) but demonstrated lower generalizability on the test dataset (R² = 0.24 vs. 0.54). The root mean square error (RMSE) for the RF model for training and testing ranged from 14 to 22 cm, compared to 17 and 18 cm for the Stefan model. Variable importance analysis revealed that mean annual temperature and slope angle were the strongest predictors of ALT, accounting for 19% and 18% of the variance, respectively, followed by sediment transport index (14%) and stream power index (11%). Comparative analysis of baseline ALT predictions showed the Stefan model tended to project a thicker active layer (mean ± SD: 65 ± 16 cm), compared to the RF model (mean ± SD: 59 ± 8.8) cm). Both models indicated a latitudinal gradient in ALT, with shallower depths at higher latitudes. Projected ALT increases by 2100 were estimated at 3.3 ± 2.2 cm under SSP 2-4.5 and 5.9 ± 4.0 cm under SSP 5-8.5 for the ML model, whereas the Stefan model projected substantially larger increases of 13 ± 2.6 cm (SSP 2-4.5) and 28 ± 4.4 cm (SSP5-8.5). Spatial analysis showed the greatest ALT increases in northern Alaska, with relatively smaller changes in southern regions. These findings highlight the complex, multifactorial nature of ALT dynamics and the value of hybrid modeling approaches. As rising temperatures accelerate permafrost thaw, changes in ALT can disrupt ecosystems, damage infrastructures, and enhance the release of stored soil carbon, highlighting the urgent need for improved predictive capabilities to inform adaptation strategies in the Arctic.

Climate sciences↗
DOE OSTI2025

A resolution independent neural operator

The Deep operator network (DeepONet) is a powerful yet simple neural operator architecture that utilizes two deep neural networks to learn mappings between infinite-dimensional function spaces. This architecture is highly flexible, allowing the evaluation of the solution field at any location within the desired domain. However, it imposes a strict constraint on the input space, requiring all input functions to be discretized at the same locations; this limits its practical applications. Here, in this work, we introduce a general framework for operator learning from input–output data with arbitrary number and locations of sensors. This begins by introducing a resolution-independent DeepONet (RI-DeepONet), enabling it to handle input functions that are arbitrarily, but sufficiently finely, discretized. To this end, we propose two dictionary learning algorithms to adaptively learn a set of appropriate continuous basis functions, parameterized as implicit neural representations (INRs), from correlated signals defined on arbitrary point cloud data. These basis functions are then used to project arbitrary input function data as a point cloud onto an embedding space (i.e., a vector space of finite dimensions) with dimensionality equal to the dictionary size, which can be directly used by DeepONet without any architectural changes. In particular, we utilize sinusoidal representation networks (SIRENs) as trainable INR basis functions. The introduced dictionary learning algorithms are then used in a similar way to learn an appropriate dictionary of basis functions for the output function data, which defines a new neural operator architecture referred to as the R esolution I ndependent N eural O perator (RINO). In the RINO, the operator learning task simplifies to learning a mapping from the coefficients of input basis functions to the coefficients of output basis functions. We demonstrate the robustness and applicability of RINO in handling arbitrarily (but sufficiently richly) sampled input and output functions during both training and inference through several numerical examples.

Deep operator network (DeepONet)↗
NASA NTRSDate not supplied

Generation of Land Surface Reflectance with Combined Geo-KOMPSAT-2A AMI and Himawari 8 AHI Observations

The latest generation of geostationary satellites has opened a new era of Earth observations with unprecedented spatiotemporal resolution and spectral range. Together with GOES 16/17 ABI, FY4-A AGRI, and Himawari-8 AHI, a new Korean geostationary satellite (Geo-KOMPSAT-2A AMI) has operationally collected a full-disk image in 16 channels every ten minutes since July 2019, allowing diurnal land surface monitoring over a large proportion of Asia and all of Oceania. Retrieving accurate surface reflectance (SR) over land from GK-2A/AMI is a challenging but high priority objective. One of the challenges is the absence of a spectral band in the 2.2 m SWIR range from AMI, which is required by many atmospheric correction algorithms to retrieve atmospheric aerosol properties. To remedy this issue, we adopt a strategy that combines concurrent GK-2A/AMI and Himawari 8/AHI observations in order to derive AMI SR. We have adapted the NASA Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm to process the data stream from Himawari 8/AHI. The advantages of the MAIAC algorithm is its capability to exploit the high temporal frequency and varying illumination geometry of the geostationary data for advanced cloud/snow detection, aerosol retrieval, and characterization of surface bidirectional reflectance properties. Leveraging the similarities of spectral bands and the sun-target-sensor geometry between AMI and AHI, we are able to create denser time series of observations and enhanced BRDF samples over most of the spatial coverage of AMI (and AHI). The combined stereo-type observations not only help derive SR for AMI but also enhance retrievals of the corresponding AHI surface products. We evaluate the resulting AMI SR using ground (AERONET) observations and corresponding MODIS products. Further, we discuss potential challenges in utilizing the geostationary satellite data for land surface monitoring.

geostationary satellite↗
NASA NTRS2006

Robust Coordination for Large Sets of Simple Rovers

The ability to coordinate sets of rovers in an unknown environment is critical to the long-term success of many of NASA;s exploration missions. Such coordination policies must have the ability to adapt in unmodeled or partially modeled domains and must be robust against environmental noise and rover failures. In addition such coordination policies must accommodate a large number of rovers, without excessive and burdensome hand-tuning. In this paper we present a distributed coordination method that addresses these issues in the domain of controlling a set of simple rovers. The application of these methods allows reliable and efficient robotic exploration in dangerous, dynamic, and previously unexplored domains. Most control policies for space missions are directly programmed by engineers or created through the use of planning tools, and are appropriate for single rover missions or missions requiring the coordination of a small number of rovers. Such methods typically require significant amounts of domain knowledge, and are difficult to scale to large numbers of rovers. The method described in this article aims to address cases where a large number of rovers need to coordinate to solve a complex time dependent problem in a noisy environment. In this approach, each rover decomposes a global utility, representing the overall goal of the system, into rover-specific utilities that properly assign credit to the rover s actions. Each rover then has the responsibility to create a control policy that maximizes its own rover-specific utility. We show a method of creating rover-utilities that are "aligned" with the global utility, such that when the rovers maximize their own utility, they also maximize the global utility. In addition we show that our method creates rover-utilities that allow the rovers to create their control policies quickly and reliably. Our distributed learning method allows large sets rovers be used unmodeled domains, while providing robustness against rover failures and changing environments. In experimental simulations we show that our method scales well with large numbers of rovers in addition to being robust against noisy sensor inputs and noisy servo control. The results show that our method is able to scale to large numbers of rovers and achieves up to 400% performance improvement over standard machine learning methods.

Tumer, Kagan↗
DOE OSTI2025

Second-generation downscaled earth system model data using generative machine learning

The second-generation Sup3rCC dataset provides high-resolution meteorological data generated through the downscaling of multiple earth system models (ESMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6). This downscaling is performed through application of a generative machine learning approach called Super-Resolution for Renewable Resource Data (sup3r). This dataset builds on the first-generation Sup3rCC data by applying improved bias correction methods and adding downscaled precipitation to the output variables. As with the first Sup3rCC version, the data still include temperature, wind speed and direction at multiple heights, pressure, three components of downwelling solar radiation, and relative humidity—all at 4-kilometer (km) hourly resolution over the contiguous United States. This is a 25x spatial enhancement and 24x temporal enhancement of the source 100-km daily-average ESM data. This extension of the Sup3rCC dataset includes data from six ESMs from two shared socioeconomic pathways (SSPs) totaling 400 years of data with multiple future projections of changing meteorological conditions. The scenario selection was based on a structured evaluation of historical ESM skill and comprehensive representation of possible trajectories of future climate change in temperature, humidity, precipitation, solar irradiance, and near-surface wind speeds. The inclusion of multiple future projections is intended to enable users to assess key drivers of un 36 certainty and variability. All data are double-bias corrected, resulting in a product that can be used out-of-the-box for energy system analysis with minimal historical bias. The potential applications of Sup3rCC data extend to various topics in renewable energy resource assessment, energy systems modeling, and grid resilience studies. High-resolution future meteorological projections are critical for evaluating the effects of changing meteorological conditions on renewable energy generation, energy demand, and for optimizing energy storage and grid infrastructure. The 4-km hourly resolution of the downscaled data enables understanding of spatial and temporal variability at the scales necessary for energy system operational planning. In addition, the dataset can support risk assessments by providing detailed information on possible future extreme weather events and long-term meteorological variability at scales relevant to energy infrastructure. By offering an enhanced representation of possible future meteorological conditions, the second-generation Sup3rCC dataset enables more precise modeling of energy resilience and adaptation strategies in response to changing meteorological conditions.

24 POWER TRANSMISSION AND DISTRIBUTION↗
NASA NTRS2018

NASA Pilot-Engaged Expert Response Using IBM Watson Technology: Prototype Evaluation of Knowledge Retrieval System

NASA Langley Research Center and IBM have been investigating the use of IBM Watson technology in aerospace research and development. One application of Watson technology is the Pilot-Engaged Expert Response (PEER) use case. The PEER system is envisioned as an in-cockpit advisor that will act as a source of situationally-relevant information for pilots and other flight crew members to assist in decision making about real-time events and situations that arise in the course of aircraft operations. PEER will make available vast stores of knowledge and information quickly and directly, putting important informational resources where they are needed most. IBM has worked with NASA to develop an architecture and articulate a roadmap for the development of the PEER system. That vision is built around Watson Discovery Advisor (WDA) software solution, derived from IBM's Jeopardy!-winning automatic question answering system. PEER makes use of WDA's sophisticated question-answering capabilities as its core, adding important User Interface components and other customizations for the cockpit environment, including communication with flight systems and other external data sources. The development plan for PEER includes four development stages, with the current project constituting the first phase. In this project, a prototype instance of PEER was successfully adapted to the aviation domain, enabling users to ask questions about aviation topics and receive useful and accurate answers to these questions. Major tasks accomplished include the development of procedures for domain adaptation through automatic lexicon extraction from domain glossaries; generation of question-answer training data which was used to train the system; and assessment of the effectiveness of domain adaptation, which showed a dramatic improvement in the ability of the PEER system to answer domain-relevant questions. In addition, the vision for the PEER system was pushed forward by the articulation of a plan for the automatic enhancement of question-answering with contextual information. This initial phase focused on two main goals: 1) the targeted domain adaptation of the underlying WDA system to the aviation domain; and, 2) the design of the software systems needed to leverage flight-contextual data. Domain adaptation of the WDA system proceeds via three main activities: Domain data ingestion, lexical customization and model training. A textual corpus consisting of 1,147 individual documents with more than 7.5 million words of text was ingested into the system and this served as the basis of all further development. A domain lexicon of over 3,500 aviation-domain terms was semi-automatically generated from domain documents and used to train the system. In addition, a set of over 500 question-answer (QA) pairs relevant to the PEER use case was developed; these were used to train and assess the system. These important first steps established the basis for the PEER system. In addition, steps were taken towards the integration of the PEER system into the cockpit environment with the development of a functional design for the Contextual Data Augmentation (CDA) subsystem. This subsystem brings to bear contextual data to improve system responses. It has three main submodules: the Contextual Data Collection module, the Contextual Data Selection module, and the Contextual QA Augmentation module. These modules form a processing pipeline that addresses the problems associated with automatically integrating information from external resources into the knowledge-retrieval mechanism.

Machine learning↗
NASA NTRS2022

A New Machine Learning Based Analysis for Improving Satellite Retrieved Atmospheric Composition Data: OMI SO2 as an Example

Despite recent progress, satellite retrievals of anthropogenic SO2 still suffer from relatively low signal-tonoise ratios. In this study, we demonstrate a new machine learning data analysis method to improve the quality of satellite SO2 products. In the absence of large ground-truth datasets for SO2, we start from SO2 slant column densities (SCDs) retrieved from the Ozone Monitoring Instrument (OMI) using a data-driven, physically based algorithm and calculate the ratio between the SCD and the root mean square (rms) of the fitting residuals for each pixel. To build the training data, we select presumably clean pixels with small SCD / rms ratios (SRRs) and set their target SCDs to zero. For polluted pixels with relatively large SRRs, we set the target to the original retrieved SCDs. We then train neural networks (NNs) to reproduce the target SCDs using predictors including SRRs for individual pixels, solar zenith, viewing zenith and phase angles, scene reflectivity, and O3 column amounts, as well as the monthly mean SRRs. For data analysis, we employ two NNs: (1) one trained daily to produce analyzed SO2 SCDs for polluted pixels each day and (2) the other trained once every month to produce analyzed SCDs for less polluted pixels for the entire month. Test results for 2005 show that our method can significantly reduce noise and artifacts over background regions. Over polluted areas, the monthly mean NN-analyzed and original SCDs generally agree to within ±15 %, indicating that our method can retain SO2 signals in the original retrievals except for large volcanic eruptions. This is further confirmed by running both the NN-analyzed and original SCDs through a topdown emission algorithm to estimate the annual SO2 emissions for ∼ 500 anthropogenic sources, with the two datasets yielding similar results. We also explore two alternative approaches to the NN-based analysis method. In one, we employ a simple linear interpolation model to analyze the original SCD retrievals. In the other, we develop a PCA–NN algorithm that uses OMI measured radiances, transformed and dimension-reduced with a principal component analysis (PCA) technique, as inputs to NNs for SO2 SCD retrievals. While the linear model and the PCA–NN algorithm can reduce retrieval noise, they both underestimate SO2 over polluted areas. Overall, the results presented here demonstrate that our new data analysis method can significantly improve the quality of existing OMI SO2 retrievals. The method can potentially be adapted for other sensors and/or species and enhance the value of satellite data in air quality research and applications.

Can Li↗
NASA NTRS2024

Artificial Intelligence Medical Support for Long-Duration Space Missions

We envision an artificial intelligence (AI) based system that will provide support and recommendations to the crew medical officer (CMO) and ground flight surgeon during long-duration space missions. Such a system would be pretrained on the knowledgebase of clinical knowledge on Earth, minimizing the amount of Earth data that needs to be transferred into space. Then during deployment, the system would be constantly refined through active learning from diverse streams of data from sensors in the spacecraft, data collected daily from individual astronauts, and human-in-the-loop feedback from the crew. The model could be interrogated for predictions and recommendations on personalized crew health based on the overall status of the spacecraft, medicinal stores, and status of other crew members. Adaptation techniques would be used to incorporate spaceflight data that have very different distributions from the training data due to the extreme environment. Edge computing and the most advanced neuromorphic processing would enable computation in scenarios with low power and bandwidth, while dimensionality reduction would be employed to ensure that the input data streams from spaceflight are as small as possible. In order to realize this long-term vision, several hardware and software aspects need to be developed and assembled. First, models pretrained on Earth biomedical data would need to be evaluated for predictive accuracy, and the best one selected. That model would need to be adapted to learn from diverse, sparse, and inconsistently measured data streams, as well as human-in-the-loop feedback. A data integration, standardization, and dimensionality reduction methodology would need to be developed to handle all data types and feed them into the model. Once the software and data infrastructure is developed, it would need to be integrated with small footprint compute processors and tested in high-radiation, high-vibration, unregulated temperature situations. As a short-term goal, we recommend to focus on the development of the data and model software structure. Several large language models (LLM) already exist that have been trained on Earth biomedical and clinical knowledgebases, including BioMedLLM, Med-PaLM, SPOKE LLM, and Foresight. These models need to be evaluated for accuracy and the best one chosen for a proof-of-concept structure, while maintaining awareness of the accelerating AI field and incorporating any newly improved model architectures as needed. Then, we recommend to develop a database of synthetic data types to mimic the diverse data streams that are expected in a long-duration space mission. This should include environmental and microbial data from the spacecraft, non-invasive data from wearables and point-of-care devices employed by astronauts, and more invasive molecular and physiological monitoring of clinical and biomarker data from astronauts. The data standardization methodology should be developed, and these data streams used to refine the clinical LLM. Several scenarios should be developed that could plausibly come up in a long-duration space mission, and changes or aberrations introduced to the data at specific times to mimic these scenarios. Then, question and answer tasks should be designed to interrogate the model for predictions and recommendations, with acceptable answers already identified.

Artificial Intelligence↗
DOE OSTI2024

An Ethics-Based Review of Generative Artificial Intelligence: Assuring Responsible Use (Version 1.0)

The rapid expansion of generative artificial intelligence (GenAI) has generated excitement regarding its potential benefits and concern over its ethical implications. Governments, corporations, and standards organizations have described ethical principles to direct GenAI's development and use; however, practical guidance for implementing these principles is limited. Addressing this gap is critical, especially considering the array of risks associated with GenAI, such as legal liabilities, privacy concerns, security threats, and potential misuse. Robust policies and procedures are critical to support responsible deployment of GenAI. This report examines Pacific Northwest National Laboratory (PNNL)’s approach to promoting responsible GenAI use. Proposed initiatives include developing policies based on ethical principles, creating a governance process to review projects relative to those principles, and implementing onboarding processes for training staff. The governance framework described in this report adapts the structure and principles of Institutional Review Boards (IRBs), traditionally used in human subjects research, for GenAI ethical review, providing oversight. Ethical principles guiding responsible GenAI usage include transparency and accountability, privacy, fairness, safety, security, and validity and reliability. To operationalize these principles, we propose forming a GenAI Assurance Council (GAC) that mirrors the IRB's structure. The GAC will evaluate GenAI projects across privacy, accountability, transparency, safety, security, fairness, and validity dimensions. Complementing policy and governance is AI literacy training to support staff understanding of GenAI's ethical implications. An initial training effort for AI Incubator Chat—a GenAI tool deployed at PNNL—showed promising results, underscoring the importance of clear guidelines and user accountability. Collaborative efforts and the dissemination of best practices are also discussed. The proposed GAC model and AI literacy training provide a blueprint for establishing ethical GenAI use and governance, offering practical tools to bridge the gap between ethical principles and real-world applications. The responsible integration of GenAI at PNNL entails a multifaceted approach involving policy development, ethical governance, and AI literacy training. The positive initial feedback and collaborative opportunities position PNNL to lead by example in GenAI's responsible use, reflecting a proactive stance in addressing the ethical, legal, and societal challenges associated with this emerging technology. PNNL's systematic and ethical approach to GenAI offers a model for other institutions to emulate, promoting safe and responsible technological advancements in the AI domain.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Large language models (LLMs) are increasingly adapted to downstream tasks via reinforcement learning (RL) methods like Group Relative Policy Optimization (GRPO), which often require thousands of rollouts to learn new tasks. We argue that the interpretable nature of language often provides a much richer learning medium for LLMs, compared to policy gradients derived from sparse, scalar rewards. To test this, we introduce GEPA (Genetic-Pareto), a prompt optimizer that thoroughly incorporates natural language reflection to learn high-level rules from trial and error. Given any AI system containing one or more LLM prompts, GEPA samples trajectories (e.g., reasoning, tool calls, and tool outputs) and reflects on them in natural language to diagnose problems, propose and test prompt updates, and combine complementary lessons from the Pareto frontier of its own attempts. As a result of GEPA's design, it can often turn even just a few rollouts into a large quality gain. Across six tasks, GEPA outperforms GRPO by 6% on average and by up to 20%, while using up to 35x fewer rollouts. GEPA also outperforms the leading prompt optimizer, MIPROv2, by over 10% (e.g., +12% accuracy on AIME-2025), and demonstrates promising results as an inference-time search strategy for code optimization. We release our code at https://github.com/gepa-ai/gepa.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

A comparative analysis of YOLOv8 and U-Net image segmentation approaches for transmission electron micrographs of polycrystalline thin films

Metallic thin films offer a platform to experimentally study the dynamics of microstructural evolution, but the required transmission electron microscopy (TEM)-based imaging generates complex images that are challenging to segment and quantify. This work provides a comparative analysis of a new YOLOv8 model and an established U-Net model for bright-field TEM images of polycrystals, employing a framework leveraging physical observables to evaluate performance against two hand-traced benchmark datasets. This methodology obviates the comparison of large, diversely structured, and manually labeled datasets that are required to assess performance on a per-image/per-pixel basis. It is found that the YOLOv8 model, adapted for real-time instance segmentation, has up to 43× faster inferencing (NVIDIA GeForce RTX 4090) compared to U-Net and reconstructs hand-traced grain size distributions (GSDs) with excellent fidelity, finding mean diameter within 3% for grains near an optimal magnification; for grains that deviate from the optimal pixel-diameter, the size of small- (large)-diameter grains is systematically over- (under)-estimated. This is partially mitigated by including scale-aware augmentations during training. Moreover, when the bias is corrected post-inference by a rigid shift in distribution, the YOLOv8 model reproduces ground truth GSDs with exceptional fidelity, with statistical tests indicating <5% probability that the distributions are distinct. Based on ground truth data, calibration curves pertaining to this shift can be constructed for a given model. This issue is not present in the U-Net model’s results, indicating that for quantitative measurements where the true size of objects is of interest, special procedures must be implemented for YOLO-based models.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Integrating Intermediate Traits in Phylogenetic Genotype-to-Phenotype Studies

A major goal of research in evolution and genetics is linking genotype to phenotype. This work could be direct, such as determining the genetic basis of a phenotype by leveraging genetic variation or divergence in a developmental, physiological, or behavioral trait. The work could also involve studying the evolutionary phenomena (e.g., reproductive isolation, adaptation, sexual dimorphism, behavior) that reveal an indirect link between genotype and a trait of interest. When the phenotype diverges across evolutionarily distinct lineages, this genotype-to-phenotype problem can be addressed using phylogenetic genotype-to-phenotype (PhyloG2P) mapping, which uses genetic signatures and convergent phenotypes on a phylogeny to infer the genetic bases of traits. The PhyloG2P approach has proven powerful in revealing key genetic changes associated with diverse traits, including the mammalian transition to marine environments and transitions between major mechanisms of photosynthesis. However, there are several intermediate traits layered in between genotype and the phenotype of interest, including but not limited to transcriptional profiles, chromatin states, protein abundances, structures, modifications, metabolites, and physiological parameters. Each intermediate trait is interesting and informative in its own right, but synthesis across data types has great promise for providing a deep, integrated, and predictive understanding of how genotypes drive phenotypic differences and convergence. We argue that an expanded PhyloG2P framework (the PhyloG2P matrix) that explicitly considers intermediate traits, and imputes those that are prohibitive to obtain, will allow a better mechanistic understanding of any trait of interest. Furthermore, this approach provides a proxy for functional validation and mechanistic understanding in organisms where laboratory manipulation is impractical.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Leveraging transfer learning and leaf spectroscopy for leaf trait prediction with broad spatial, species, and temporal applicability

Accurate and reliable prediction of leaf traits is crucial for understanding plant adaptations to environmental variation, monitoring terrestrial ecosystems, and enhancing comprehension of functional diversity and ecosystem functioning. Currently, various approaches (e.g., statistical, physical models) have been developed to estimate leaf traits through hyperspectral remote sensing and leaf spectroscopy. However, the absence of high-performing, transferable, and stable models across various domains of space, plant functional types (PFTs) and seasons hinder our ability to quantify and comprehend spatiotemporal variations in leaf traits. This study proposes robust and highly transferable models for better predicting leaf traits with hyperspectral reflectance. Initially, three datasets were assembled, pairing common leaf traits — chlorophyll (Chla+b), carotenoids (Ccar), leaf mass per area (LAM), equivalent water thickness (EWT) — with leaf spectra measurements collected across diverse geographic locations in the U.S. and Europe, PFTs, and seasons. Measurements were acquired using spectroradiometers (e.g., ASD FieldSpec 3/4/Pro and SVC HR-1024i) with integrating spheres, leaf clips, and contact probes. Here, we then developed transfer learning-based hybrid models that incorporated the domain knowledge of radiative transfer models (RTMs) through pretraining processes and were well-constrained by fine-tuning with field measurements. Through comparison with other state-of-the-art statistical models, including partial-least squares regression (PLSR) and Gaussian Process Regression (GPR), as well as pure physical models, we found that the proposed transfer learning models achieved better predictive performance and higher transferability. Specifically, compared to other statistical models and pure RTMs, the transfer learning model exhibited higher coefficient of determination (R 2 ) values with range of 0.01 to 0.79, lower normalized root mean square error (NRMSE) with range of 0.06 % to 33.25 % in model performance. Additionally, the models exhibited improved transferability, with higher R 2 values range from 0.04 to 0.32, lower NRMSE range from 0.08 % to 30.81 %. The findings underscore that transfer learning models through integrating domain knowledge from RTMs and limited observations, can harness the advantages of both RTMs and statistical models and serve as a promising approach for effectively predicting leaf traits.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

scPlantAnnotate: an accurate and robust transformer-based model for plant cell type annotation

Accurate cell type annotation remains a major bottleneck in plant single-cell RNA sequencing (scRNA-seq), where existing tools are often adapted from animal studies and perform sub-optimally on plant data. The lack of plant-specific computational frameworks limits the construction of plant cell atlases and downstream biological discovery. We develop and evaluate scPlantAnnotate, a Transformer-based reference annotation framework tailored for plant scRNA-seq data, and benchmark it against state-of-the-art deep learning and conventional methods across multiple plant species. Species-specific scPlantAnnotate models were trained using curated datasets from Arabidopsis thaliana, Zea mays, Oryza sativa, and Glycine max. We compared scPlantAnnotate with leading baselines under both standard random-split evaluation and a more stringent leave-one-dataset-out setting, which tests robustness to completely unseen datasets and tissue types. scPlantAnnotate consistently outperforms existing approaches across all four species under random-split evaluation. In the leave-one-dataset-out setting for A. thaliana, where performance drops markedly for all methods due to strong batch effects and dataset heterogeneity, scPlantAnnotate nonetheless achieves the highest Accuracy, Macro-F1, Balanced Accuracy, and Macro-AUROC on average and ranks first on most held-out datasets. These results demonstrate improved robustness to dataset shifts, a critical yet underexplored challenge in plant scRNA-seq analysis. A freely accessible web server enables users to annotate their own datasets using pretrained models. scPlantAnnotate provides a plant-specific, Transformer-based framework for single-cell annotation that delivers state-of-the-art performance and enhanced robustness to unseen datasets. By addressing limitations of existing tools and enabling scalable reference-based annotation, scPlantAnnotate supports the development of comprehensive plant cell atlases and facilitates broader use of single-cell genomics in plant biology.

Bioinformatics↗
DOE OSTI2024

Hydrogen Detection Strategies to Support H2@SCALE - The NREL Sensor Laboratory

Hydrogen represents a major pathway to decarbonize and stabilize the national and international energy industry and select manufacturing markets. To facilitate the development of hydrogen markets, the US Department of Energy initiated H2@Scale to bring together stakeholders to advance affordable hydrogen production, transport, storage, and utilization to increase revenue opportunities in multiple energy sectors. One major impediment to hydrogen implementation is cost. To expedite the use of hydrogen in energy and other markets, the United States announced in 2021 the Hydrogen Shot, which seeks to reduce the cost of clean hydrogen by 80% to $1 per 1 kilogram in 1 decade ("1 1 1"). As the cost of hydrogen drops, new applications will emerge that will require unique configurations of existing equipment and infrastructure, and eventually lead to advances in the generation and utilization of hydrogen. As the hydrogen economy expands, sensors and detection methods will need to adapt to changing infrastructure demands to address the primary targets of health & safety, emissions monitoring, and process control. The NREL Sensor Laboratory is playing a pivotal role in advancing the use of hydrogen sensors and detection methodologies in each of these categories to support DOE's mission for safe and efficient utilization in emerging markets. Health & safety monitors are required to ensure that operators and facilities can react to unintended hydrogen releases, either as GH2, LH2, or as a constituent of blends (e.g., natural gas or ammonia). Current detection methodologies focus on safety applications to detect near its lower flammable limit (4 vol %), and typically include point sensors in applications such as fixed or mobile detectors (e.g., personal gas monitors). Methodologies amenable for area detection include acoustic, emerging optical imaging methods, and flame detectors. Comparable detection strategies can be utilized for emissions monitoring and quantization, however few methods can simultaneously cover both low (emissions) and high (health & safety) levels. Deployment of emission level detectors will be required to 1) reduce product loss through small but potentially significant leaks from an environmental or cost perspective, 2) reduce downtime of high demand systems by early identification of eminent system failures (leaks through pump or compressor seals indicative of impending failure), and 3) address potential emission monitoring requirements that may be set by regulating bodies. The first two points should be adopted by industry to reduce the cost-of-goods-sold. The third main category for hydrogen detection relates to process control and may be advantageous for many existing applications. Two main applications are emerging. For example, the purity requirements for hydrogen that is dispensed from refueling systems for hydrogen fuel cell electric vehicles (FCEV) is rigorously regulated by the Standard SAE J2719, which prescribes maximum allowable levels of multiple impurities in the hydrogen fuel and must be verified by a regulatory body. Hydrogen contaminant detectors (HCD) integrated to the fueling station can assure this compliance. HCDs must be able operate in 100% H2 backgrounds and be able to distinguish between multiple contaminants at low ppm to low ppb levels. Secondly, as a strategy to decarbonize the natural gas grid, there are proposals to blend hydrogen with natural gas. This blending will affect transport applications (pipeline infrastructure), stationary combustion systems (turbines), and consumer and commercial appliances. In the short-term, hydrogen levels up to 20% are proposed. Variations in the hydrogen level can have dramatic impact on the combustion process and on the potential response of safety sensors. These mixtures may be regulated so that the concentration at a delivery point must be monitored with high precision. However, routine maintenance may introduce background gases such as ambient air (with water) or maintenance gases (introduced with welding processes or adhesive outgassing.) Therefore, the detection methodology must be robust enough to recover or respond to various contaminants. Several reviews can be found in literature addressing sensing and detection technologies, including their limitations and applications. However, for most applications, limitations can be alleviated by combining various detection techniques either through system integration or implementation of machine learning methods (artificial intelligence). In this presentation, we will discuss several applications, highlight their current approach for hydrogen detection, and suggest detection strategies to supplement their limitations.

ENERGY STORAGE,HYDROGEN↗