Results for “Robust artificial intelligence” — Page 10

DOE OSTI2025

Cosmological parameter estimation with a joint-likelihood analysis of the cosmic microwave background and big bang nucleosynthesis

Here, we present a joint-likelihood analysis of big bang nucleosynthesis (BBN) and cosmic microwave background (CMB) data, consistently combining likelihoods and taking into account uncertainties in nuclear reaction rates for the first time. Bayesian inference is performed on the baryon abundance and the effective number of neutrino species, 𝑁 eff , using a CMB Boltzmann solver in combination with LINX , a new flexible and efficient BBN code. We marginalize over Planck nuisance parameters and nuclear rates to find 𝑁 eff =3.0⁢8$^{+0.15}_{−0.14}$, 2.9⁢4$^{+0.16}_{−0.15}$, or 2.96$^{+0.13}_{−0.14}$, for three separate reaction networks. This framework enables robust testing of the lambda cold dark matter paradigm and its variants with CMB and BBN data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

NASA NTRSDate not supplied

Towards an Aviation Large Language Model by Fine-tuning and Evaluating Transformers

In the aviation domain, there are many applications for machine learning and artificial intelligence tools that utilize natural language. For example, there is a desire to know the commonalities in written safety reports such as voluntary post incidents reports or create more accurate transcripts of air traffic management conversations. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement (LOA) which is used as the evaluation case in this paper. These applications can benefit from the use of state-of-the-art Natural Language Processing (NLP) techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of transferring pre-trained large language models to the aviation domain by adapting transformer based models using aviation datasets. This paper utilized two datasets to adapt a ‘Robustly Optimized Bidirectional Encoder Representations from Transformers Approach’ (RoBERTa) model and two down-stream classification tasks to assess its performance. These datasets are all built upon Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. The first two datasets are used for the adaptation of RoBERTa to the aviation domain and were of different sizes to assess the number of documents needed to adapt to the aviation domain. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the unsupervised adaptation. The second dataset is a separate set of LOA documents with two sets of classification labels to be used for evaluation; one at the document level and one at the line level. These down-stream evaluations allowed the measurement of improvement by adapting RoBERTa. The accuracy increased by 4-6% on both tasks and the F1 score on the class of interest increased by 4-8% from the adaptation.

Air Traffic Management↗

NASA NTRSDate not supplied

Towards an Aviation Large Language Model by Fine-tuning and Evaluating Transformers

In the aviation domain, there are many applications for machine learning and artificial intelligence tools that utilize natural language. For example, there is a desire to know the commonalities in written safety reports such as voluntary post incidents reports or create more accurate transcripts of air traffic management conversations. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement (LOA) which is used as the evaluation case in this paper. These applications can benefit from the use of state-of-the-art Natural Language Processing (NLP) techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of transferring pre-trained large language models to the aviation domain by adapting transformer based models using aviation datasets. This paper utilized two datasets to adapt a ‘Robustly Optimized Bidirectional Encoder Representations from Transformers Approach’ (RoBERTa) model and two down-stream classification tasks to assess its performance. These datasets are all built upon Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. The first two datasets are used for the adaptation of RoBERTa to the aviation domain and were of different sizes to assess the number of documents needed to adapt to the aviation domain. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the unsupervised adaptation. The second dataset is a separate set of LOA documents with two sets of classification labels to be used for evaluation; one at the document level and one at the line level. These down-stream evaluations allowed the measurement of improvement by adapting RoBERTa. The accuracy increased by 4-6% on both tasks and the F1 score on the class of interest increased by 4-8% from the adaptation.

Air Traffic Management↗

DOE OSTI2026

Plant Design for a Developing Bioeconomy Workshop Report: Frontier Science for the Bioeconomy Workshop Series

Recent advances in fundamental plant biology research, synthetic biology, and artificial intelligence (AI) are unlocking powerful new capabilities in plant biodesign, offering unprecedented potential to reimagine plants as programmable platforms for resource-efficient production of bioenergy, biomaterials, chemicals, and more. The U.S. Department of Energy (DOE) convened the Plant Design for a Developing Bioeconomy virtual workshop on March 12 through 14, 2025, to bring together leaders across plant science, engineering, and computation to assess the current landscape and define a bold vision for future research. Discussions during the workshop built upon findings included in DOE’s Biological and Environmental Research (BER) workshop report Overcoming Barriers in Plant Transformation: A Focus on Bioenergy Crops (U.S. DOE 2024; genomicscience. energy.gov/plant-transformation). Participants identified critical knowledge gaps, technical barriers, and emerging opportunities in the design and engineering of plant systems to support a robust, resilient domestic bioeconomy aligned with DOE’s mission.

09 BIOMASS FUELS↗

DOE OSTI2026

Multi‐Model Ensembles in Ecosystem Modeling: Challenges and Best Practices for Decision‐Making

Ecosystem models are increasingly central to the decision-making for environmental policy, conservation planning, and climate-related investments. Yet, the growing reliance on Multi-Model Ensembles (MMEs) of ecosystem models by practitioners and policymakers, sometimes under tight timelines and imperfect information, has frequently outpaced the scientific rigor required to ensure ensemble reliability. Here, MMEs refer to approaches that combine targeted predictions from multiple models with the expectation of improving robustness and quantifying predictive uncertainty. Poorly designed MMEs may create a false sense of confidence and lead to suboptimal policy and market decisions. This perspective argues that robust decision-making-relevant MMEs must be grounded on two pillars: (1) rigorous Model Intercomparison Projects (MIPs), which identify inter-model agreement and disagreement, characterize model uncertainties, and evaluate robustness with observationally based benchmarks—MIPs' diagnostic evaluation is so critical that it must be needed to drive MME's decision in model selection and weighting, especially when only a limited number of models available; and (2) co-design by both stakeholders and scientists to ensure that scenarios, metrics and uncertainty requirements provide decision-relevant information. Building upon the past success and lessons from the existing MIPs-MMEs efforts (e.g., climate/Earth system/crop), we derived the theoretical basis for MMEs, addressed their specific challenges in ecosystem modeling, and highlighted proper consideration of model numbers and diversity, risk of model inter-dependence, effective calibration of model parameters, possible overdue of some ecosystem model development, critical roles of open benchmark data across a wide range of conditions, and suggested use of Artificial Intelligence to support MIPs-MMEs. We highlighted the under-recognized opportunity for MIPs and MMEs to drive scientific progress and innovation through identifying better performing models, systematic benchmarking, feedback loops, and targeted model improvement. By following actionable best practice guidelines, MMEs can evolve from ad hoc aggregation of models into a trusted backbone of environmental policy and decision-making.

ecosystem modeling↗

DOE OSTI2020

State-of-the-art on research and applications of machine learning in the building life cycle

Fueled by big data, powerful and affordable computing resources, and advanced algorithms, machine learning has been explored and applied to buildings research for the past decades and has demonstrated its potential to enhance building performance. This study systematically surveyed how machine learning has been applied at different stages of building life cycle. By conducting a literature search on the Web of Knowledge platform, we found 9579 papers in this field and selected 153 papers for an in-depth review. The number of published papers is increasing year by year, with a focus on building design, operation, and control. However, no study was found using machine learning in building commissioning. There are successful pilot studies on fault detection and diagnosis of HVAC equipment and systems, load prediction, energy baseline estimate, load shape clustering, occupancy prediction, and learning occupant behaviors and energy use patterns. None of the existing studies were adopted broadly by the building industry, due to common challenges including (1) lack of large scale labeled data to train and validate the model, (2) lack of model transferability, which limits a model trained with one data-rich building to be used in another building with limited data, (3) lack of strong justification of costs and benefits of deploying machine learning, and (4) the performance might not be reliable and robust for the stated goals, as the method might work for some buildings but could not be generalized to others. Finally, findings from the study can inform future machine learning research to improve occupant comfort, energy efficiency, demand flexibility, and resilience of buildings, as well as to inspire young researchers in the field to explore multidisciplinary approaches that integrate building science, computing science, data science, and social science.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

NASA NTRSDate not supplied

NASA R&M Efficiency through Findable, Accessible, Interoperable, and Reusable (FAIR) Digital Assets

At the intersection of mission, technology, and place is NASA’s need to modernize for a digital-forward future. Digitalization, the process of moving toward digital business, is occurring everywhere and remains an ongoing process across the federal government.”[1] Whereas, Digital Transformation is “employing digitization/digital technologies (e.g., Artificial Intelligence (AI), mobile, cloud, data) to change a process, product, or capability so dramatically (e.g., real-time, intelligent, personalized, anywhere, anytime) that it is unrecognizable compared to its traditional form.” [2] In order to facilitate a digital transformation it is essential for NASA to understand and identify where data exists today and which data are value-needed in the future, understand where there are unfulfilled data needs that limit the advancement of NASA work, and ensure NASA efficiency through Findable, Accessible, Interoperable, and Reusable (FAIR) digital assets in the future. Therefore, NASA’s Reliability & Maintainability (R&M) Enterprise Data Sharing team is working to leverage both Digitization and Digital Transformation to achieve their vision of developing an R&M data discovery framework that enables our community, our partners, and our stakeholders with the ability to efficiently, robustly, and seamlessly access information that enables real-time knowledge and model-based, analytics driven, decision-making impacting R&M. As a result the R&M Enterprise Data Sharing team has conducted a survey of its Reliability, Maintainability, and Availability (RMA) community members to identify data existence (created or used) and where there are corresponding barriers to data acquisition and/or R&M or other issues as shown within this presentation.

Digital Transformation↗

NASA NTRSDate not supplied

Trilateral Task Force – Reliability Analysis Supporting Mission Extension/Post Mission Disposal

At the intersection of mission, technology, and place is NASA’s need to modernize for a digital-forward future. Digitalization, the process of moving toward digital business, is occurring everywhere and remains an ongoing process across the federal government.”[1] Whereas, Digital Transformation is “employing digitization/digital technologies (e.g., Artificial Intelligence (AI), mobile, cloud, data) to change a process, product, or capability so dramatically (e.g., real-time, intelligent, personalized, anywhere, anytime) that it is unrecognizable compared to its traditional form.” [2] In order to facilitate a digital transformation it is essential for NASA to understand and identify where data exists today and which data are value-needed in the future, understand where there are unfulfilled data needs that limit the advancement of NASA work, and ensure NASA efficiency through Findable, Accessible, Interoperable, and Reusable (FAIR) digital assets in the future. Therefore, NASA’s Reliability & Maintainability (R&M) Enterprise Data Sharing team is working to leverage both Digitization and Digital Transformation to achieve their vision of developing an R&M data discovery framework that enables our community, our partners, and our stakeholders with the ability to efficiently, robustly, and seamlessly access information that enables real-time knowledge and model-based, analytics driven, decision-making impacting R&M. As a result the R&M Enterprise Data Sharing team has conducted a survey of its Reliability, Maintainability, and Availability (RMA) community members to identify data existence (created or used) and where there are corresponding barriers to data acquisition and/or R&M or other issues as shown within this presentation.

Digital Transformation, Reliability Engineering↗

DOE OSTI2025

Artificial intelligence time series forecasting for feed-forward laser stabilization

Laser plasma accelerators, typically operating at 1–10 Hz repetition rates, have the ability to produce high-quality electron beams in compact, all-optical-driven configurations, with the electron beams uniquely suited for a wide variety of accelerator-based applications. However, fluctuations and drifts in the laser delivery to the meter-scaled and below plasma target (the electron beam source) will translate into electron beam source variations that can limit their utility for demanding applications like light sources or linear colliders. Commercially available active feedback laser stabilization systems are intrinsically bandwidth limited due to their integration with multi-inch corrective mirror mounts which minimizes their effectiveness. In this manuscript, we present a Neural Network time series forecaster that can predict laser position fluctuations of the laser delivery to the final target well ahead of a future laser shot. The Root-Mean-Square-Error (RMSE) of the prediction accuracy was < 2 μ m for a 1 / e 2 beam radius of 34 μ m . Our feed-forward approach serves as a first-step in circumventing the bandwidth limitations imposed by the currently available stabilization systems since it allows for mirrors to be moved into position ahead of time to offset the predicted future position drift. This will help advance laser plasma accelerator research by providing greater robustness and stability needed for its applications.

Berger, Curtis↗

DOE OSTI2026

Beta-Delayed Neutron Emission of 𝑁=84 132Cd

Using the time-of-flight technique, we measured the beta-delayed neutron emission of 132 Cd . From our large-scale shell model (LSSM) calculation using the N3⁢LO interaction [, Phys. Rev. Lett. 131, 022501 (2023)], we suggest the decay is dominated by the transformation of a neutron in the 𝑔7/2 orbital, deep below the Fermi surface, into a proton in the 𝑔9/2 orbital. We compare the beta-decay half-lives and neutron branching ratios of nuclei with 𝑍<50 and 𝑁 ≥82 obtained with our LSSM with those of leading “global” models such as finite-range droplet model (FRDM). Our calculations match known half-lives and neutron branching ratios well and suggest that current leading models overestimate the yet-to-be-measured half-lives. Our model, backed by the 132 Cd decay data presented here, offers robust predictive power for nuclei of astrophysical interest such as 𝑟-process waiting points.

Madurga, M [Department of Physics and Astronomy, U↗

DOE OSTI2025

Data Quality Monitoring for the Hadron Calorimeters Using Transfer Learning for Anomaly Detection

The proliferation of sensors brings an immense volume of spatio-temporal (ST) data in many domains, including monitoring, diagnostics, and prognostics applications. Data curation is a time-consuming process for a large volume of data, making it challenging and expensive to deploy data analytics platforms in new environments. Transfer learning (TL) mechanisms promise to mitigate data sparsity and model complexity by utilizing pre-trained models for a new task. Despite the triumph of TL in fields like computer vision and natural language processing, efforts on complex ST models for anomaly detection (AD) applications are limited. In this study, we present the potential of TL within the context of high-dimensional ST AD with a hybrid autoencoder architecture, incorporating convolutional, graph, and recurrent neural networks. Motivated by the need for improved model accuracy and robustness, particularly in scenarios with limited training data on systems with thousands of sensors, this research investigates the transferability of models trained on different sections of the Hadron Calorimeter of the Compact Muon Solenoid experiment at CERN. The key contributions of the study include exploring TL’s potential and limitations within the context of encoder and decoder networks, revealing insights into model initialization and training configurations that enhance performance while substantially reducing trainable parameters and mitigating data contamination effects.

47 OTHER INSTRUMENTATION↗

DOE OSTI2024

Efficient Mixed-Precision Matrix Factorization of the Inverse Overlap Matrix in Electronic Structure Calculations with AI-Hardware and GPUs

In recent years, a new kind of accelerated hardware has gained popularity in the artificial intelligence (AI) community which enables extremely high-performance tensor contractions in reduced precision for deep neural network calculations. In this article, we exploit Nvidia Tensor cores, a prototypical example of such AI-hardware, to develop a mixed precision approach for computing a dense matrix factorization of the inverse overlap matrix in electronic structure theory, S –1 . This factorization of S –1 , written as ZZT = S –1 , is used to transform the general matrix eigenvalue problem into a standard matrix eigenvalue problem. Here we present a mixed precision iterative refinement algorithm where Z is given recursively using matrix–matrix multiplications and can be computed with high performance on Tensor cores. To understand the performance and accuracy of Tensor cores, comparisons are made to GPU-only implementations in single and double precision. Additionally, we propose a nonparametric stopping criteria which is robust in the face of lower precision floating point operations. The algorithm is particularly useful when we have a good initial guess to Z, for example, from previous time steps in quantum-mechanical molecular dynamics simulations or from a previous iteration in a geometry optimization.

36 MATERIALS SCIENCE↗

DOE OSTI2026

Artificial Intelligence Thermostat to Detect Faults

Residential air conditioners and heat pumps often experience faults due to inadequate maintenance, which can severely reduce efficiency or even cause system failure. Common issues include dirty or clogged air filters and refrigerant leaks. These problems degrade performance and increase energy use and operating costs. This study presents a smart thermostat with embedded artificial intelligence to detect such faults and alert homeowners when maintenance is needed. The thermostat uses low-cost measurements—including return-air temperature, relative humidity, supply-air temperature, outdoor-air temperature, and condenser subcooling—to identify abnormal operations. Because different faults produce distinct response patterns, tailored algorithms are developed to recognize characteristic fault signatures. The investigation is built on a detailed co-simulation platform that couples EnergyPlus with the DOE/ORNL Heat Pump Design Model (HPDM). EnergyPlus represents the building’s dynamic environment, while HPDM is a high-fidelity, hardware-based model that can simulate fault-free performance as well as a wide range of faults, including gradual degradation such as minor refrigerant leakage. This platform provides a virtual training and testing environment that helps distinguish fault-induced behavior from normal operation and supports development of robust diagnostic algorithms. Using this framework, a Dynamic Bayesian Network was developed to identify two common faults—gradual refrigerant charge loss and indoor airflow blockage—and the AI-embedded thermostat was verified through annual building simulations.

Shen, Bo [ORNL] (ORCID:0000000336600393)↗

DOE OSTI2020

Leveraging Flexible Smart Manufacturing to Accelerate Industrial Supply Chain Recovery

Any major crisis such as the latest coronavirus disease pandemic will have a monumental effect on the worldwide economy and on the international supply chain. Much of the global supply chain relies on China, Germany, and the United States for manufacturing and distribution; this fragile system is subject to failure. Climactic changes in demand and rapid shortages in necessities are the results of the pandemic’s disruption. The recovery forecast for the ongoing pandemic is uncertain and, therefore, any recovery effort will need to be adaptable to build the supply chain resiliency. A wide adoption of flexible smart automated technologies in the manufacturing sector are helping to build robust supply chains and assisting in recovery. Here, a brief discussion of these smart manufacturing technologies is presented, along with a list of potential supply chain issues and corresponding solutions. Such smart manufacturing technologies possess the potential to empower and revolutionize the traditional manufacturing environment by enhancing its resiliency and flexibility. These smart technologies will play a crucial role in accelerating the worldwide supply chain recovery in and after a pandemic.

99 GENERAL AND MISCELLANEOUS↗

NASA NTRS1995

Applying AI tools to operational space environmental analysis

The U.S. Air Force and National Oceanic Atmospheric Agency (NOAA) space environmental operations centers are facing increasingly complex challenges meeting the needs of their growing user community. These centers provide current space environmental information and short term forecasts of geomagnetic activity. Recent advances in modeling and data access have provided sophisticated tools for making accurate and timely forecasts, but have introduced new problems associated with handling and analyzing large quantities of complex data. AI (Artificial Intelligence) techniques have been considered as potential solutions to some of these problems. Fielding AI systems has proven more difficult than expected, in part because of operational constraints. Using systems which have been demonstrated successfully in the operational environment will provide a basis for a useful data fusion and analysis capability. Our approach uses a general purpose AI system already in operational use within the military intelligence community, called the Temporal Analysis System (TAS). TAS is an operational suite of tools supporting data processing, data visualization, historical analysis, situation assessment and predictive analysis. TAS includes expert system tools to analyze incoming events for indications of particular situations and predicts future activity. The expert system operates on a knowledge base of temporal patterns encoded using a knowledge representation called Temporal Transition Models (TTM's) and an event database maintained by the other TAS tools. The system also includes a robust knowledge acquisition and maintenance tool for creating TTM's using a graphical specification language. The ability to manipulate TTM's in a graphical format gives non-computer specialists an intuitive way of accessing and editing the knowledge base. To support space environmental analyses, we used TAS's ability to define domain specific event analysis abstractions. The prototype system defines events covering reports of natural phenomena such as solar flares, bursts, geomagnetic storms, and five others pertinent to space environmental analysis. With our preliminary event definitions we experimented with TAS's support for temporal pattern analysis using X-ray flare and geomagnetic storm forecasts as case studies. We are currently working on a framework for integrating advanced graphics and space environmental models into this analytical environment.

Krajnak, Mike↗

DOE OSTI2026

Accelerated Selectrion of Optimal Perovskite Alloys for Solar PV using a Combined Quantum and Machine Learning Hierachiral Approach

The project aims to: (i) accelerate the discovery of “Missing HP alloys” by combining quantum mechanics and artificial intelligence machine learning approaches, and (ii) analyze the stabilities of candidate alloys, including those that do not pass selection filters (and are thus expected to degrade over time) to decipher the nature of the instabilities to guide the development of durable solar cell materials. Successful candidates will be subjected to validation experiments at NREL's state-of-the-art facilities. The discoveries this effort will provide will be directly testable and implementable and will greatly impact U.S. progress in HP PV as they will provide clear direction and motivation for experimental studies including specific material synthetic targets, device optimization, and device stability protocols. A key advantage of this effort is the feedback and guidance provided by the Industry Collaborative Work Group that we established to coordinate academic and national lab research with industry needs. The proposed work will provide a basis for and direct the development of robust and reliable HP PV. It will also provide a timely, valuable and extensive roadmap to the experimental HP PV community to enable it to focus its efforts on improving and fine-tuning promising HP compositions that this effort predicts will likely be the best performers rather than wandering in the vast chemical space for decades spending enormous resources mostly evaluating unpromising candidate materials.

14 SOLAR ENERGY↗

NASA NTRS1991

Acquisition and improvement of human motor skills: Learning through observation and practice

Skilled movement is an integral part of the human existence. A better understanding of motor skills and their development is a prerequisite to the construction of truly flexible intelligent agents. We present MAEANDER, a computational model of human motor behavior, that uniformly addresses both the acquisition of skills through observation and the improvement of skills through practice. MAEANDER consists of a sensory-effector interface, a memory of movements, and a set of performance and learning mechanisms that let it recognize and generate motor skills. The system initially acquires such skills by observing movements performed by another agent and constructing a concept hierarchy. Given a stored motor skill in memory, MAEANDER will cause an effector to behave appropriately. All learning involves changing the hierarchical memory of skill concepts to more closely correspond to either observed experience or to desired behaviors. We evaluated MAEANDER empirically with respect to how well it acquires and improves both artificial movement types and handwritten script letters from the alphabet. We also evaluate MAEANDER as a psychological model by comparing its behavior to robust phenomena in humans and by considering the richness of the predictions it makes.

Iba, Wayne↗

DOE OSTI2025

Vision Foundation Models in Remote Sensing: A survey

Artificial intelligence (AI) technologies have profoundly transformed the field of remote sensing (RS), revolutionizing data collection, processing, and analysis. Traditionally reliant on manual interpretation and task-specific models, RS research has been significantly enhanced by the advent of foundation models (FMs)—large-scale pretrained AI models capable of performing a wide array of tasks with unprecedented accuracy and efficiency. This article provides a comprehensive survey of FMs in the RS domain. We categorize these models based on their architectures, pretraining datasets, and methodologies. Through detailed performance comparisons, we highlight emerging trends and the significant advancements achieved by those FMs. Additionally, we discuss technical challenges, practical implications, and future research directions, addressing the need for high-quality data, computational resources, and improved model generalization. Our research also finds that pretraining methods, particularly self-supervised learning (SSL) techniques like contrastive learning (CL) and masked autoencoders (MAEs), remarkably enhance the performance and robustness of FMs. This survey aims to serve as a resource for researchers and practitioners by providing a panorama of advances and promising pathways for the continued development and application of FMs in RS.

data models↗