Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model reduction”

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 91 records · Page 5
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Interpreting and Accelerating Transformers for Jet Tagging

Attention-based transformers are ubiquitous in machine learning applications from natural language processing to computer vision. In high energy physics, one central application is to classify collimated particle showers in colliders based on the particle of origin, known as jet tagging. In this work, we study the interpretatbility and prospects for acceleration of Particle Transformer (ParT), a state-of-the-art model, leverages particle-level attention to improve jet-tagging performance. We analyzing ParT's attention maps and particle-pair correlations in the eta-phi plane, revealing intriguing features, such as a binary attention pattern that identifies critical substructure in jets. These insights enhance our understanding of the model's internal workings and learning process and hint at ways to improve its efficiency. Along these lines, we also explore low-rank attention, attention alternatives, and dynamic quantization to accelerate transformers for jet tagging. With quantization, we achieve a 50% reduction in model size and a 10% increase in inference speed without compromising accuracy. These combined efforts enhance both the performance and the interpretability of transformers in high-energy physics, opening avenues for more efficient and physics-driven model designs.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Understanding and Estimating Error Propagation in Neural Networks for Scientific Data Analysis

Neural networks are increasingly integrated into scientific discovery, where input data reduction and model quantization play a key role in accelerating inference. However, understanding and mitigating the impact of these techniques on output error is critical for ensuring reliable results, particularly in tasks demanding high numerical precision. This paper introduces a comprehensive framework for optimizing neural network inference in scientific computing by combining data reduction and weight quantization while maintaining error-controlled outcomes. We develop theoretical analyses to bound error propagation under these reductions and propose a framework that balances computational performance with error constraints. Evaluation on real-world learning-based combustion simulations and satellite image classification demonstrates that our derived error bounds accurately predict observed errors while enabling significant computational speedup under our framework. This work highlights the potential for further leveraging advancements in modern lossy compression algorithms and hardware accelerators that support lower-precision formats.

He, Weiming [New Jersey Institute of Technology]↗
DOE OSTI2024

PV Performance Modeling and Stakeholder Engagement (Final Technical Report)

This core capability project’s objective is to increase the value of photovoltaic (PV) performance models by improving their functionality, demonstrating, and quantifying their validity, and offering a wide range of stakeholder engagement opportunities. In FY22-24, we developed new and improved modeling algorithms and functions to represent PV performance more accurately in a variety of environments and conditions. The “Model parameter toolkit” was developed and includes functions to translate between different module temperature models, incidence angle modifier models, and single-diode models. A new modeling capability named “PV Atlas” was also developed leveraging Sandia’s High Performance Computing resources. This capability allows us to investigate several questions and provide climate-specific best practices and geographic data files; all these are hosted on an interactive website on Sandia’s GitHub and can be used for training, system optimization, or to provide best practices for uncertainty reduction. For model validation, we published high-quality PV performance, and weather data; these data are well documented, filtered, and processed for quality and include examples on how to run PV simulations. We also developed well documented, standardized methods for validating PV models and ran independent model validation and 2 blind modeling intercomparisons engaging with 49 organizations from 17 countries. We co-led and contributed to a growing, well documented and maintained suite of open-source functions for PV modeling (i.e., the pvlib-python) and we outreached to the PV modeling stakeholders via the PVPMC workshops and web resources. In addition, this project supported US representation and leadership for the International Energy Agency (IEA) PVPS Task 13; specifically, members of our team led and supported 3 subtasks on: 1) Best practices for the optimization of bifacial photovoltaic tracking, 2) Extreme weather events and their multiple impact on PV power plants: Risks, failure mechanisms and mitigation strategies, and 3) Best practice guidelines for the use of economic and technical Key Performance Indicators (KPIs). This project resulted in the publications of 14 peer reviewed journal papers, 37 conference presentations, 6 SAND reports, 5 public datasets and 6 new webpages on the PVPMC website. It supported the release of 13 pvlib-python versions where 28 enhancements were from this PV Performance Modeling project. We co-organized 5 PVPMC workshops in FY22-24 with the participation of 214 unique institutions and around 700 participants. The PVPMC website was redesigned, and its reliability was improved; it receives over 50,000 visitors/year from 202 unique countries.

14 SOLAR ENERGY↗
DOE OSTI2025

Impact of a dynamic grid mix and climate on operational carbon emissions modeling for different building typologies and climate zones

Calculating operational carbon emissions through a building’s lifecycle is complex due to the dynamic nature of influencing factors such as climate and energy grid mix. This paper introduces a novel methodology for modeling 30-year operational carbon impacts of buildings and applies this method to mid-rise office and residential typologies across various US climate zones. The method accounts for these temporal variabilities using new and scarcely cited data sources. Key findings indicate that future changes in the climate, while impactful, play a relatively modest role in operational carbon emissions compared to significant reductions with modeling scenarios using the projected decarbonization of the electricity grid. Here, the study also finds that using annual, month-hourly, or hourly grid emission factors have a minimal impact on carbon accounting, except in certain climates and program types where emission patterns do not align with a building’s energy consumption. Warmer climates like Miami, Florida and Tucson, Arizona, which rely heavily on cooling, demonstrate larger variations in carbon emissions when using higher temporal resolution emission factors. Ultimately, this study underscores the critical role of grid decarbonization in reducing long-term emissions and the importance of incorporating this variable in life cycle assessment (LCA) modeling.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Factors That Influence Variability in Stress-Drop Measurements Using Spectral Decomposition and Spectral-Ratio Methods for the 2019 Ridgecrest Earthquake Sequence

Stress drop is a fundamental parameter related to earthquake source physics, but is hard to measure accurately. To better understand how different factors influence stress-drop measurements, we compare two different methods using the Ridgecrest stress-drop validation data set: spectral decomposition (SD) and spectral ratio (SR), each with different processing options. Here, we also examine the influence of spectral complexity on source parameter measurement. Applying the SD method, we find that frequency bandwidth and time-window length could influence spectral magnitude calibration, while depth-dependent attenuation is important to correctly map stress-drop variations. For the SR method, we find that the selected source model has limited influence on the measurements; however, the Boatwright model tends to produce smaller standard deviation and larger magnitude dependence than the Brune model. Variance reduction threshold, frequency bandwidth, and time-window length, if chosen within an appropriate parameter range, have limited influence on source parameter measurement. For both methods, wave type, attenuation correction, and spectral complexity strongly influence the result. The scale factor that quantifies the magnitude dependence of stress drop show large variations with different processing options, and earthquakes with complex source spectra deviating from the Brune-type source models tend to have larger scale factor than earthquakes without complexity. Based on these detailed comparisons, we make a few specific suggestions for data processing workflows that could help future studies of source parameters and interpretations.

58 GEOSCIENCES↗
DOE OSTI2026

Forest residue harvest optimization: spanning the bridge between plant biology and biorefinery performance

Forestry residues have immense potential as alternative feedstocks to petroleum, yet their inherent complexity remains a major challenge to widespread use. Pairing the temporal rhythms of plant biology with biorefinery performance is critical to industrial-scale biorefinery development. Here, we provide the first report of a techno-economic analysis (TEA) and life cycle assessment (LCA) for a model integrated reductive catalytic fractionation (RCF)–molten salt hydrolysis process for forestry residues varying in tree part, species, and phenophase. All forestry residues resulted in net-negative greenhouse gas (GHG) emissions vs. comparable petroleum feedstocks, with GHG emissions potentially reduced >4.0× through composition-based feedstock selection (e.g., harvesting American beech bark in spring vs. summer). Moreover, American beech twigs/branchlets and bark in leafed and emergence phenophases, respectively, had 7.9× lower predicted phenolic minimum selling prices (MSPs) vs. other feedstocks and MSPs within the current global phenolic market range. Hemicellulose content and RCF yield emerged as key parameters impacting GHG emissions and biorefinery revenue, identifying hardwood twigs/branchlets in the leafed phenophase as optimal biofeedstocks. Biorefinery expenses were dominated by purchased equipment, raw materials, and utility costs, highlighting essential areas for future study. Notably, RCF reactor pressures drove 85–90% of equipment costs, but sensitivity analysis revealed that decreasing the pressure 20% could reduce the phenol MSP 4-fold. Structural carbohydrate dynamics were also investigated using a two-step acid hydrolysis method to resolve tissue- and species-level patterns in biomass composition throughout the year to enable harvest optimization based on TEA/LCA findings. Ultimately, elucidating the impact of biofeedstock dynamics on biorefinery performance enables harvest optimization, informed engineering design, and progress towards an expanded bioeconomy.

Shapiro, Alison J. [University of Delaware, Newark↗
DOE OSTI2024

Event Detection and Classification Using Machine Learning Applied to PMU Data for the Western US Power System

Smart grid technology enhances our comprehension and reliability of the power grid, leveraging Phasor Measurement Unit (PMU) data—time-synchronized, high-frequency measurements gathered across the US power grid. This paper employs machine learning techniques to effectively analyze the vast PMU data in Wide Area Monitoring Systems (WAMS) for power grid event detection and classification. Analyzing several months of real-world PMU data, the paper focuses on machine learning for fast, precise event detection and classification, corroborated by utility event logs. Practical challenges like feature extraction, dimensionality reduction, and model selection are addressed. A novel feature yielding improved results is discovered, and a supplementary algorithm for detecting small power grid faults is developed. The final algorithm is validated using a month-long real PMU data set, demonstrating its capability in accurately identifying power grid events in near real-time.

machine learning, event detection, PMU↗
DOE OSTI2024

Efficiency of ML Anomaly Detection Triggers for Emerging Jets

Novel machine learning-based anomaly detection Level 1 (L1) triggers are currently under development at CMS, namely AXOL1TL and CICADA. The former employs a variational autoencoder, while the latter utilizes a convolutional autoencoder. These triggers aim to balance rate reduction with model independence, enabling the selection of potentially significant events that might be overlooked by traditional triggers relying on basic kinematic variable selections. Consequently, they have the potential to enhance signals indicative of physics beyond the Standard Model, such as those associated with emerging jets. Such signals are predicted by models featuring a composite dark sector where long-lived particles decay into Standard Model jets with displaced tracks and numerous vertices. This study evaluates the efficiency of these anomaly detection triggers in selecting events with emerging jets produced via the s-channel production of two dark quarks.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

EFIT‐AI: Machine Learning and Artificial Intelligence Assisted Equilibrium Reconstruction for Tokamak Experiments and Burning Plasmas (Final Report)

The EFIT-AI project is creating a modern advanced equilibrium reconstruction code suitable for tokamak experiments of burning plasmas. EFIT [1,2] was the first and is the most extensively used equilibrium reconstruction code in the world. This project builds on the production-level experience and adds key elements as follows. 1. A Model Order Reduction (MOR) version of the two-dimensional (2D) Grad-Shafranov equation solver (EFIT-MORNN) using physics-informed neural networks. 2. Improved optimization and data analysis capabilities using a Bayesian framework enhanced with machine learning. 3. A MOR version of the three-dimensional (3D) perturbed equilibrium reconstruction tool.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

CMIP7 data request: land and land ice priorities and opportunities

The Land and Land Ice Theme in the Coupled Model Intercomparison Project Phase 7 (CMIP7) represents the current understanding of physical processes in land surface ecosystems, hydrology, cryosphere, and their physical interactions with other Earth system components. Simulations from Earth system models (ESMs) could provide crucial information for assessing planetary safety, such as critical tipping elements, and be used to inform climate risks for improving climate impact assessments and policy decisions. This paper presents a collaborative effort to identify scientific opportunities in the Land and Land Ice Theme of the CMIP7 Data Request. The proposed opportunities build upon advances in ESMs, including new freshwater system and land ice processes being included in CMIP7, as well as the scientific community's demand for high-frequency and sub-grid-scale land surface outputs. In total, 25 variable groups that contain 716 variables have been identified to be potentially available to the broad scientific audience for performing analysis in land–atmosphere coupling, hydrological processes and freshwater systems, glacier and ice sheet mass balance and their influence on the sea levels, land use, and plant phenology. Key reflections from this data request effort include advocacy for closer engagement between the user community and modeling groups, reduction in the technical barriers to tracking existing parameters and defining new variables, and more streamlined variable management. These will be essential to enhance the usability and reliability of CMIP7 outputs for climate and Earth system research and applications to a broad audience that relies on the CMIP7 endeavor.

Li, Yue [Univ. of California, Los Angeles, CA (Uni↗
DOE OSTI2024

Efficiency of ML Anomaly Detection Triggers for Emerging Jets

Novel machine learning-based anomaly detection Level 1 (L1) triggers are currently under development at CMS, namely AXOL1TL and CICADA. The former employs a variational autoencoder, while the latter utilizes a convolutional autoencoder. These triggers aim to balance rate reduction with model independence, enabling the selection of potentially significant events that might be overlooked by traditional triggers relying on basic kinematic variable selections. Consequently, they have the potential to enhance signals indicative of physics beyond the Standard Model, such as those associated with emerging jets. Such signals are predicted by models featuring a composite dark sector where long-lived particles decay into Standard Model jets with displaced tracks and numerous vertices. This study evaluates the efficiency of these anomaly detection triggers in selecting events with emerging jets produced via the s-channel production of two dark quarks.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Efficiency of ML Anomaly Detection Triggers for Emerging Jets

Novel machine learning-based anomaly detection Level 1 (L1) triggers are currently under development at CMS, namely AXOL1TL and CICADA. The former employs a variational autoencoder, while the latter utilizes a convolutional autoencoder. These triggers aim to balance rate reduction with model independence, enabling the selection of potentially significant events that might be overlooked by traditional triggers relying on basic kinematic variable selections. Consequently, they have the potential to enhance signals indicative of physics beyond the Standard Model, such as those associated with emerging jets. Such signals are predicted by models featuring a composite dark sector where long-lived particles decay into Standard Model jets with displaced tracks and numerous vertices. This study evaluates the efficiency of these anomaly detection triggers in selecting events with emerging jets produced via the s-channel production of two dark quarks.

43 PARTICLE ACCELERATORS↗
DOE OSTI2024

ML-based Dimension Reduction Strategies

Deep learning (DL)--based surrogate models have achieved success in various applications in carbon capture and storage (CCS). However, the model training on high-dimensional spaces is computationally expensive and impractical for large-scale and complex geological models, because the models usually contain hundreds of thousands to millions of grid cells, each with a set of parameters. Furthermore, the high cost of generating training data with sufficient variation is another limitation of model training on high-dimensional spaces, which may result in overfitting and reduce the model efficiency and prediction performance. We proposed the workflow incorporating dimension reduction methods and deep learning models, which aim to extract the latent variables of input parameters and output state variables, and then build the mapping function at the latent spaces. The proposed workflow can significantly reduce the computational complexity in solving both forward and inverse problems compared to models trained on high-dimensional spaces. Dimensionality reduction models showed great potential in workflows for fast reservoir simulation, history matching, prior model generation, visualization, and more, ultimately enhancing DL model performance in related SMART Work Packages.

Hosseini, Seyyed↗
DOE OSTI2024

Machine learning models of intermittent operation of RO wellhead water treatment for salinity reduction and nitrate removal

Machine learning models were developed for intermittent multi-mode operation of a wellhead reverse osmosis water purification and desalination system to predict salt passage, nitrate passage, and permeate flux. The models, based on long short-term memory (LSTM) recurrent neural network (RNN) architecture, included an attention mechanism to increase model performance in proximity of the regulatory limit for nitrate. Training and testing of the models for the Startup, Production, Shutdown and Flushing operational modes were based on operational data (consisting of 22 process variables per data sample) acquired every 2–5 s over a six-month period. The significant sets of model input attributes for the different operational modes were assessed via Spearman ranking correlation, Self-Organizing Map (SOM) analysis and feed forward feature selection (FFFS). Although the variability of nitrate passage, salt passage and permeate flux was significant over the four operational modes, prediction performance for the three outcomes were with R2 and Average Absolute Relative Error (AARE) of 0.78–0.95 and 2.96–6.16 %, respectively. Model updates post membrane elements replacement demonstrated similar levels of prediction accuracy. The study results suggest that there is merit in exploring the utility of multi-mode models for sensor fault detection, data imputation, and for potential use in model-predictive control.

Intermittent RO operation↗
DOE OSTI2025

Decoding diffraction and spectroscopy data with machine learning: A tutorial

This Tutorial provides a step-by-step guide on how to apply supervised machine-learning techniques to analyze diffraction and spectroscopy data. This Tutorial details four models—a reconstruction-focused model, a regression-focused model, a hybrid reconstruction/regression model, and a multimodal model—that use x-ray diffraction profiles and vibrational density of states spectra to predict various microstructural descriptors. In this Tutorial, we cover data pre-processing steps, constructions of the models via dimensionality reduction and regression, training, and analysis of these models. Comparisons of the model’s performance are provided, highlighting the strength and weakness of the various approaches utilized.

36 MATERIALS SCIENCE↗
DOE OSTI2024

High-Fidelity Arc-Discharge Model for Hydrogen-Plasma-Smelting-Reduction of Iron Ore

Electrification and use of renewable hydrogen is currently a necessity for decarbonizing the iron-and-steel industry. In this regard, hydrogen plasma smelting reduction (HPSR) is a novel pathway that is being explored for reduction of iron ore. HPSR provides several decarbonization merits compared to conventional blast furnaces. Firstly, the use of renewable hydrogen drastically reduces the CO2 emissions compared to the use of coke. Secondly, renewable electricity in the form of a thermal plasma for making reactive hydrogen species (radicals, ions) are more efficient at reducing iron ore compared to neutral H2. Thirdly, a molten product compatible with downstream processes is obtained from the intense heat transfer from the plasma. However, the scale-up of this technology requires fundamental exploration of hydrogen plasma dynamics and its interaction with complex solid material that include phase changing iron-ore and slag. In this work, we present a first principles continuum scale model for thermal plasmas in Ar/H2 gas mixtures typically used for HPSR. The thermal plasma governing equations for mass, momentum and energy with Lorentz force and Joule heating source terms are solved along with electromagnetic equations for electrostatic and magnetic vector potential. Our solver will be based on Pele, a suite of reacting flow solvers designed for advanced scientific computing architectures (Henry De Frahan et al., Proceedings of SIAM Parallel Processing, 13-25, 2024), and will utilize adaptive mesh generation for enhanced resolutions at locations of intense physicochemical interactions. This study will present the impact of Ar to H2 ratios on excited/dissociated hydrogen species concentrations, plasma temperature and conductivity along with the impact of outgassed species (water, metal vapor, O, OH radicals) from ore surface on gas phase chemistry. Furthermore, the heat and species flux to the surface will be quantified as a function of applied voltages in a transferred arc configuration.

hydrogen plasma↗
DOE OSTI2025

An evaluation of multi-fidelity methods for quantifying uncertainty in projections of ice-sheet mass change

Abstract. This study investigated the computational benefits of using multi-fidelity statistical estimation (MFSE) algorithms to quantify uncertainty in the mass change of Humboldt Glacier, Greenland, between 2007 and 2100 using a single climate change scenario. The goal of this study was to determine whether MFSE can use multiple models of varying cost and accuracy to reduce the computational cost of estimating the mean and variance of the projected mass change of a glacier. The problem size and complexity were chosen to reflect the challenges posed by future continental-scale studies while still facilitating a computationally feasible investigation of MFSE methods. When quantifying uncertainty introduced by a high-dimensional parameterization of the basal friction field, MFSE was able to reduce the mean-squared error in the estimates of the statistics by well over an order of magnitude when compared to a single-fidelity approach that only used the highest-fidelity model. This significant reduction in computational cost was achieved despite the low-fidelity models used being incapable of capturing the local features of the ice-flow fields predicted by the high-fidelity model. The MFSE algorithms were able to effectively leverage the high correlation between each model's predictions of mass change, which all responded similarly to perturbations in the model inputs. Consequently, our results suggest that MFSE could be highly useful for reducing the cost of computing continental-scale probabilistic projections of sea-level rise due to ice-sheet mass change.

54 ENVIRONMENTAL SCIENCES↗