SEARCH · Engineering Papers

Results for “evaluation datasets”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 685 records · Page 38

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Alaska's Rural Building Stock: a Validation Study Using ResStock and Field Data

The availability of accurate national data on demographics, building stock, and energy use is vital for modeling residential buildings and evaluating decarbonization strategies. However, rural and Indigenous populations, including those in rural Alaska, are typically underrepresented in these datasets. These communities face unique challenges due to their remote locations, severe weather conditions, and limited access to resources, resulting in high energy burden. This report examines how rural Alaskan communities are underrepresented in the ResStock housing model and highlights the need for improved data to address their unique housing and energy challenges. Thus, this report examines the representation of rural Alaskan communities within the national housing stock model, ResStock. A validation study was conducted, considering ResStock, Field Data and Aerial and 3D-view data collection (A3DDC) datasets. The validation process started by using the down selecting approach on the ResStock building stock dataset. For the purpose of this study, only the rural Alaska Boroughs and Census areas located in ASHRAE IECC Climate Zone 8 were considered to ensure a more accurate and fair comparison with the field data, which was collected in rural areas located in climate zone 8, specifically within the Nome Census area. While ResStock may accurately represent several characteristics of the building stock for rural Alaska, some differences between modeled, field data, and aerial and 3D-view data collection datasets were identified. The following building characteristics have a high impact on modeled energy consumption and demonstrated large differences: Revisit heating setpoints and consider a substantially higher setpoint distribution, it could potentially address "missing loads" if this is the case. Develop and include Toyo heating in future modeling for ResStock and EnergyPlus. Remove natural gas as a water heater fuel type outside of North Slope County. Foundation type updated to have more crawlspaces rather than basements. Infiltration rates need reexamination for a larger distribution toward higher infiltration rates. Include more vinyl and less brick in exterior wall type and revise wall color for greater proportion of light rather than dark color. Roof material revised from majority shingles to majority metal. Update number of occupants to higher number of occupant count. Building orientation represents a higher proportion of south facing buildings rather than relatively equal. The findings suggest that updating ResStock's probability logic could better represent rural Alaskan buildings. ResStock can be utilized to identify the best upgrades or energy efficiency and energy efficiency improvements, helping community leaders in making more informed decisions.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2023

Machine-Learning-Based Rotating Detonation Engine Diagnostics: Evaluation for Application in Experimental Facilities

Real-time monitoring of combustion behavior is a crucial step toward actively controlled rotating detonation engine (RDE) operation in laboratory and industrial environments. Various machine learning methods have been developed to advance diagnostic efficiencies from conventional postprocessing efforts to real-time methods. Here this work evaluates and compares conventional techniques alongside convolutional neural network (CNN) architectures trained in previous studies, including image classification, object detection, and time series classification, according to metrics affecting diagnostic feasibility, external applicability, and performance. Real-time, capable diagnostics are deployed and evaluated using an altered experimental setup. Image-based CNNs are applied to externally provided images to approximate dataset restrictions. Image classification using high-speed chemiluminescence images and time series classification using high-speed flame ionization and pressure measurements achieve classification speeds enabling real-time diagnostic capabilities, averaging laboratory-deployed diagnostic feedback rates of 4–5 Hz. Object detection achieves the most refined resolution of 20 μs in postprocessing. Image and time series classification require the additional correlation of sensor data, extending their time-step resolutions to 80 ms. Comparisons show that no single diagnostic approach outperforms its competitors across all metrics. This finding justifies the need for a machine learning portfolio containing a host of networks to address specific needs throughout the RDE research community.

33 ADVANCED PROPULSION SYSTEMS↗

DOE OSTI2026

Farm Practice Typologies as a Strategy for Management-Relevant Land Use and Land Cover Mapping in the Great Lakes Region (Version 1) [Dataset]

Dataset overview and development This dataset provides spatially explicit agricultural land-use and land-management typologies developed for the Great Lakes Region (GLR) at the farm-parcel level. The typologies were designed to characterize not only the land-use and land-cover (LULC) associated with individual agricultural farm parcels, but also the land-management practices (LMPs), including irrigation, tile drainage, and conservation easements, occurring within those parcels and how these characteristics change through time. The dataset contains four related typology products: Annual integrated typology – describes the combined LULC and land-management characteristics for each farm parcel for individual years. LULC transition typology – describes the temporal pattern of LULC change for each farm parcel across the study period (2008-2023). LMP trend typology – describes the temporal pattern in the occurrence of LMPs for each farm parcel across the study period. Multi-year integrated typology – combines the LULC transition typology and LMP trend typology to provide an integrated characterization of long-term land-use and management patterns. Purpose of the dataset The purpose of these products is to provide a management-relevant integrated and consistent framework for evaluating the spatial and temporal organization of agricultural landscapes across the GLR. The resulting typologies can: support landscape-scale environmental and land-use analysis; provide spatial information relevant to land-management strategies, conservation planning, policy development, and program evaluation; characterize spatial patterns of agricultural land use and management; examine changes in agricultural landscapes through time; and identify persistent, transitional, and changing agricultural systems. Please refer to the README file provided in Files for more details.

Agriculture↗

DOE OSTI2024

Characterization of Precipitation-Strengthening Heat-Resistant Austenitic Stainless Steels for Life-Prediction Modeling

In this study, the role of minor alloying additions in 347H stainless steels (UNS34709, ASTM A240/240M) on creep-rupture properties at 650-750°C and microstructure evolution during isothermal exposure at 750°C has been investigated, aiming to provide the experimental dataset as boundary conditions of physics-based modeling for material/component life prediction. Four different 347H heats containing various amounts of boron and nitrogen additions were prepared and evaluated. The combined additions of B and N are found to stabilize the strengthening secondary M23C6 carbides and retarding the transition from M23C6 to sigma phase precipitates during thermal exposure. The observed kinetics of microstructure evolution reasonably explains the improvement of creep-rupture properties of 347H stainless steels with the B and N additions.

Yamamoto, Yukinori↗

DOE OSTI2022

Speed vs Accuracy: Effect on Ligand Pose Accuracy of Varying Box Size and Exhaustiveness in AutoDock Vina

Abstract Structure‐based virtual high‐throughput screening involves docking chemical libraries to targets of interest. A parameter pertinent to the accuracy of the resulting pose is the root mean square deviation (RMSD) from a known crystallographic structure, i. e., the ‘docking power’. Here, using a popular algorithm, Autodock Vina, as a model program, we evaluate the effects of varying two common docking parameters: the box size (the size of docking search space) and the exhaustiveness of the global search (the number of independent runs starting from random ligand conformations) on the RMSD from the PDBbind v2017 refined dataset of experimental protein‐ligand complexes. Although it is clear that exhaustiveness is an important parameter, there is wide variation in the values used, with variation between 1 and >100. We, therefore, evaluated a combination of cubic boxes of different sizes and five exhaustiveness values (1, 8, 25, 50, 75, 100) within the range of those commonly adopted. The results show that the default exhaustiveness value of 8 performs well overall for most box sizes. In contrast, for all box sizes, but particularly for large boxes, an exhaustiveness value of 1 led to significantly higher median RMSD (mRMSD) values. The docking power was slightly improved with an exhaustiveness of 25, but the mRMSD changes little with values higher than 25. Therefore, although low exhaustiveness is computationally faster, the results are more likely to be far from reality, and, conversely, values >25 led to little improvement at the expense of computational resources. Overall, we recommend users to use at least the default exhaustiveness value of 8 for virtual screening calculations.

Agarwal, Rupesh↗

DOE OSTI2020

A novel improved model for building energy consumption prediction based on model integration

Building energy consumption prediction plays an irreplaceable role in energy planning, management, and conservation. Constantly improving the performance of prediction models is the key to ensuring the efficient operation of energy systems. Moreover, accuracy is no longer the only factor in revealing model performance, it is more important to evaluate the model from multiple perspectives, considering the characteristics of engineering applications. Based on the idea of model integration, this paper proposes a novel improved integration model (stacking model) that can be used to forecast building energy consumption. The stacking model combines advantages of various base prediction algorithms and forms them into “meta-features” to ensure that the final model can observe datasets from different spatial and structural angles. Two cases are used to demonstrate practical engineering applications of the stacking model. A comparative analysis is performed to evaluate the prediction performance of the stacking model in contrast with existing well-known prediction models including Random Forest, Gradient Boosted Decision Tree, Extreme Gradient Boosting, Support Vector Machine, and K-Nearest Neighbor. The results indicate that the stacking method achieves better performance than other models, regarding accuracy (improvement of 9.5%–31.6% for Case A and 16.2%–49.4% for Case B), generalization (improvement of 6.7%–29.5% for Case A and 7.1%-34.6% for Case B), and robustness (improvement of 1.5%–34.1% for Case A and 1.8%–19.3% for Case B). The proposed model enriches the diversity of algorithm libraries of empirical models.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2020

Structure–activity relationship-based chemical classification of highly imbalanced Tox21 datasets

Abstract The specificity of toxicant-target biomolecule interactions lends to the very imbalanced nature of many toxicity datasets, causing poor performance in Structure–Activity Relationship (SAR)-based chemical classification. Undersampling and oversampling are representative techniques for handling such an imbalance challenge. However, removing inactive chemical compound instances from the majority class using an undersampling technique can result in information loss, whereas increasing active toxicant instances in the minority class by interpolation tends to introduce artificial minority instances that often cross into the majority class space, giving rise to class overlapping and a higher false prediction rate. In this study, in order to improve the prediction accuracy of imbalanced learning, we employed SMOTEENN, a combination of Synthetic Minority Over-sampling Technique (SMOTE) and Edited Nearest Neighbor (ENN) algorithms, to oversample the minority class by creating synthetic samples, followed by cleaning the mislabeled instances. We chose the highly imbalanced Tox21 dataset, which consisted of 12 in vitro bioassays for > 10,000 chemicals that were distributed unevenly between binary classes. With Random Forest (RF) as the base classifier and bagging as the ensemble strategy, we applied four hybrid learning methods, i.e., RF without imbalance handling (RF), RF with Random Undersampling (RUS), RF with SMOTE (SMO), and RF with SMOTEENN (SMN). The performance of the four learning methods was compared using nine evaluation metrics, among which F 1 score, Matthews correlation coefficient and Brier score provided a more consistent assessment of the overall performance across the 12 datasets. The Friedman’s aligned ranks test and the subsequent Bergmann-Hommel post hoc test showed that SMN significantly outperformed the other three methods. We also found that a strong negative correlation existed between the prediction accuracy and the imbalance ratio (IR), which is defined as the number of inactive compounds divided by the number of active compounds. SMN became less effective when IR exceeded a certain threshold (e.g., > 28). The ability to separate the few active compounds from the vast amounts of inactive ones is of great importance in computational toxicology. This work demonstrates that the performance of SAR-based, imbalanced chemical toxicity classification can be significantly improved through the use of data rebalancing.

Idakwo, Gabriel↗

DOE OSTI2023

Biological particles and aerosol-cloud interactions in the Southern Great Plains (Final Report)

The United States Southern Great Plains (SGP) is the genesis for much of the warm season precipitation in the central and eastern United States. While atmospheric thermodynamics and large-scale dynamics play an important role in formation of precipitation, precipitation mechanisms are also sensitive to aerosols. Aerosols can suppress precipitation as cloud condensation nuclei (CCN) or act as ice nucleating particles (INP) in deep convective clouds. This project focused on understanding the role of primary biological aerosol particles (PBAP) in the region and its influence on cloud formation. Specifically, we focused on biological aerosol in the form of pollen, one type of PBAP that is emitted in large yet variable quantities from vegetation in the mid latitudes. Field observations provide evidence of pollen in the planetary boundary layer and pollen components in cloud droplets and fine particulate matter. Further, pollen grains can easily rupture when wet, forming smaller, sub pollen particles with sizes less than one micron. We evaluated the potential for PBAP events at the Department of Energy (DoE) SGP Atmospheric Radiation Measurement (ARM) research facility due to the rich dataset available. We also developed model simulations that accounted for pollen emission and the generation of sub pollen particles, which have been shown to be both cloud condensation nuclei and ice nucleating particles. The proposed work was designed to address the following questions: 1. What are the physical and chemical signatures of biological aerosol such as pollen in SGP ARM measurements? 2. What is the role of pollen-derived particles on deep convection and precipitation in the Central US? Using measurements from the DoE SGP ARM site and recent airborne campaigns, we evaluated the signatures of pollen and pollen-derived aerosols on optical properties, cloud properties and precipitation (Subba et al., 2021). We identified pollen-driven events over the data record and used the Weather Research and Forecasting Model with fully coupled chemistry (WRF-Chem) to conduct chemically realistic simulations of aerosols during and summer mesoscale convective events in the Southern Great Plains (Subba et al., in review). This work improved our understanding of the role of biological aerosol on clouds and precipitation in the Central United States and placed these results in context with anthropogenically-driven processes.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Biological particles and aerosol-cloud interactions in the Southern Great Plains

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

A cross-dimensional analysis of data-driven short-term load forecasting methods with large-scale smart meter data

Electricity load forecasting is essential to utility operation and power grid stability. A wide spectrum of data-driven methods, ranging from linear regression models to more recent deep learning models have been adopted to forecast electric load over the years. However, there still lacks a holistic evaluation of the applicability of conventional statistical and machine learning based algorithms with respect to different temporal and spatial scopes, computational requirements, and sensitivity of model-tuning. Enabled by a large-scale electricity load profile dataset of over 40,000 residential customers in a utility region, we conducted a cross-dimensional analysis of data-driven load forecasting methods. Three regression-based and seven deep learning algorithms with different model configurations were evaluated in terms of their overall and peak load prediction accuracy, and training burdens, across spatial aggregation levels ranging from the transformer, feeder, substation, to neighborhood. We found, first, the load forecasting accuracy is constrained by a predictability boundary, influenced by the forecasting horizon and spatial aggregation level. Specifically, RandomForest, XGBoost, TFT, TSMixer, and TiDE models achieved less than 10 % prediction error for up to 96-h ahead forecasting for district, substation, and feeder levels, while other models struggle at long-horizon predictions; Second, for winter and summer peak load dates, most models were able to predict the peak demand timing within ± 1 h, but the prediction percentage error varied by models, with TFT and TiDE models being the top performers; Third, models with similar prediction accuracy can differ in training burden by an order of magnitude. Therefore, choosing model configurations that balance prediction performance and computational resource is an important practical consideration for large-scale deployment of the machine learning based load forecasting. The outcome of this study can guide researchers and practitioners to choose the proper load forecasting algorithms based on their problem scope, required accuracy, and available resources. The predictability boundary can serve as a benchmark for electricity load forecasting problems with new algorithms and datasets.

Li, Han↗

DOE OSTI2023

PRIMED for the Future: Purposing Raw Intake for Machine Learning-Enabled Detection (Final Report)

The COVID-19 pandemic demonstrated how a novel, elusive, and diffuse biological threat can engender uncertainty and misinformation, and it underscored the need for flexible analytical modalities agnostic to the identity of biological material. Yet even before the pandemic, recognition of the limitations of the current, list-based approach, which focuses on known pathogens and biotoxins, and of the importance of agent-agnostic biodetection was growing within the biosecurity community. In a 2018 report on “Biodefense in the Age of Synthetic Biology,” for example, the National Academy of Sciences stated that “an overreliance on the Select Agent List is a systematic weakness affecting many aspects of the United States’ current biodefense mitigation capability”. More recently, a group of biodefense researchers proposed the identification and adoption of “bioagent-agnostic signatures (BASs)” as a way of detecting and characterizing not only existing agents but also novel ones, an approach they believe will “enable a more flexible and resilient biodefense posture”. Indeed, the future of biodetection requires us to begin developing novel analytics that can identify anomalies and/or characteristics that indicate a potential threat, whether known or unknown, without looking for a specific signature that has been identified previously. To assess potential threats more rapidly, it is critical to develop agnostic artificial intelligence (AI)/machine learning (ML) systems that can be employed for real-time assessment of the nature and source of a perturbation. Such systems should be multi scale and multi-dimensional, integrating sensor data from a range of biological, chemical, and physical application spaces. Emerging deep learning (DL) models demonstrate exceptional promise for identification of discriminatory features within multi-dimensional datasets. DL models have the capacity to recognize and encode highly complex patterns in a wide range of input data modalities, including images, text, and biological/chemical/physical spectra. As such, they can execute a wide range of assessments and determinations that have traditionally required a human operator. The promise of advances in DL is apparent in the realm of human health and medicine. DL models have been validated for evaluating a variety of clinical threats to human health in a range of contexts, including infection and cancer, and they demonstrated improved performance in predicting stroke relative to human neurologists in some categories of data. Continuously evolving advances in AI/ML are expected to support more efficient evaluation of raw sequence, spectroscopy, and spectrometry data. For instance, recent advances and deployment of large language models (LLM) such as Generative Pre training Transformer (GPT) and Bidirectional Encoder Representations from Transformers (BERT) have already motivated application of these models for biological function prediction. As frameworks such as LLMs become larger and more complex in their representations, their capacity to serve as pre-trained models that can be fine-tuned for biological/biodetection purposes will similarly be amplified. While existing and emerging AI/ML have found broad applicability and use cases in the clinical sciences, development for environmental evaluation and biodetection has been limited. Functionalizing such capabilities for this purpose requires an understanding of the existing technical landscape and how the respective tools and algorithms are currently being employed. This landscape awareness then allows an assessment of the current practical capabilities of existing models and the anticipated requirements and development efforts that will be needed to adapt available algorithms for biodetection applications relevant to DHS. Leveraging expertise in biodetection, ML, and operational biodetection, the effort described in this report is comprised of a systematic landscape assessment (Subtask 2.1), comparative evaluation (Subtask 2.2), and formulation of a value proposition (Subtask 2.3) for the prospect of ML-enabled, agnostic biodetection from raw, or minimally-processed, datasets.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

An Evaluation and Qualification of U.S.-Based Research Reactors for Irradiation Capabilities Supporting Advanced Nuclear Systems

Irradiation experiments are a prerequisite for evaluating nuclear reactor system designs, analyzing the performance of these systems, and obtaining licenses. Likewise, irradiation facilities are necessary for producing the radioisotopes used in industrial and medical applications. Recent developments in modeling and simulation capabilities and advancements in computational resources have further enabled the design of irradiation experiments for evaluating radiation-induced phenomena and determining nuclear fuel, material, and system design and safety criteria pertaining to both normal and accident scenarios. These computational tools and models require comprehensive experimental datasets acquired under prototypic radiation conditions—for exploring material and system performance under the uniquely harsh environments found in nuclear reactors—to enable verification and validation for qualification and licensing purposes. However, qualification of irradiation experimental facilities, primarily research and test reactors (RTRs), necessitates that their performance be evaluated based on the irradiation environment (e.g. flux, power, testing capabilities) using an appropriate scoring matrix. Although many university campus RTRs are available for research and development (R&D) activities and initiatives, this study focuses on evaluating and qualifying the irradiation facilities (mostly RTRs) within the United States that are suitable for advanced nuclear fuel, material, and system irradiation experiments aimed at establishing operational-performance limits and informing component and fuel designs so as to improve operational efficiencies and mitigate proliferation vulnerabilities, as well as for radioisotope production aimed at multipurpose applications. As a result, the findings of the present study support the acceleration of nuclear fuel and material qualifications, thus hastening new and advanced nuclear energy system demonstrations and radioisotope production efforts by using extended R&D.

irradiation experiment↗

DOE OSTI2025

TropiRoot 1.0: Database of tropical root characteristics across environments

Tropical ecosystems contain the world's largest biodiversity of vascular plants. Yet, our understanding of tropical functional diversity and its contribution to global diversity patterns is constrained by data availability. This discrepancy underscores an urgent need to bridge data gaps by incorporating comprehensive tropical root data into global datasets. Here, we provide a database of tropical root characteristics. This new database, TropiRoot 1.0, will be instrumental in evaluating an array of hypotheses pertaining to root functional ecology and plant biogeography, both within the tropics and relative to other global biomes. The data compilation was conducted by the TropiRoot Initiative, in partnership with the Fine-Root Ecology Database (FRED) and the Global Root Trait (GRooT) database, Colorado State University (CSU) and the Smithsonian Tropical Research Institute (STRI). Literature search and data extraction were conducted between 2020 and 2024. Literature was identified using Web of Science, Scopus, and complemented using the expert knowledge of members of TropiRoot. To provide broad environmental and geographical distributions, literature searches included root characteristics (traits) across global change drivers, natural gradients, and from different continents. We adopted FRED standardized data columns and streamlined the format to enhance accessibility for data extraction across various user groups. This optimized framework resulted in a smaller, yet comprehensive datasheet. To make the database compatible with other global root trait initiatives, column identification was standardized following the codes provided by FRED. These efforts culminated in data extracted from 104 new sources, resulting in more than 8000 rows of data (either species or community data). Most of the data in TropiRoot 1.0 include root characteristics such as root biomass, morphology, root dynamics, mass fraction, architecture, anatomy, physiology, and root chemistry. This initiative represents a 30% increase in the currently available data for tropical roots in FRED. TropiRoot 1.0 contains root characteristics from 25 different countries, where seven are located in Asia, six in South America, five in Central America and the Caribbean, four in Africa, two in North America, and 1 in Oceania. Due to the volume of data, when ancillary data were available, including soil data, these data were either extracted and included in the database or its availability was recorded in an additional column. Multiple contributors checked the entries for outliers during the collation process to ensure data quality. For text-based observations, we examined all cells to ensure that their content relates to their specific categories. For numerical observations, we ordered each numerical value from least to greatest and plotted the values, checking apparent outliers against the data in their respective sources and correcting or removing incorrect or impossible values. Some data (soil and aboveground) have different columns for the same variable presented in different units, including originally published units, but root characteristics data had units converted to match those reported in FRED. By filling a gap from global databases, TropiRoot 1.0 expands our knowledge of otherwise so far underrepresented regions and our ability to assess global trends. This advancement can be used to improve tropical forest representation in vegetation models. The data are freely available and should be cited when used.

FRED↗

DOE OSTI2020

Air compressor load forecasting using artificial neural network

Air compressor systems are responsible for approximately 10% of the electricity consumed in United States and European Union industry. As many researches have proven the effectiveness of using Artificial Neural Network inair compressor performance prediction, there is still a need to forecast the air compressor electrical load profile. The objective of this study is to predict compressed air systems’ electrical load profile, which is valuable to industry practitioners as well as software providers in developing better practice and tools for load managementand look-ahead scheduling programs. Two artificial neural networks, Two-Layer Feed-Forward Neural Networkand Long Short-Term Memory were used to predict an air compressors electrical load. Compressors with three different control mechanisms are evaluated with a total number of 11,874 observations. Here, the forecasts were validated using out-of-sample datasets with 5-fold cross-validation. Models produced average coefficient ofdetermination values from 0.24 to 0.94, average root-mean-square errors from 0.05 kW - 5.83 kW, and meanabsolute scaled errors from 0.20 to 1.33. The results indicate that both artificial neural networks yield goodresults for compressors using variable speed drive (average R 2 = 0.8 and no naïve forecasting), only the longshort-term memory model gives acceptable results for compressors using on/off control (average R 2 = 0.82 and no naïve forecasting), and no satisfactory results are obtained for load/unload type air compressors (models constituting active forecasting).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2022

Machine learning prediction on the fractional free volume of polymer membranes

Fractional free volume (FFV) characterizes the microstructural level features of polymers and affects their properties including thermal, mechanical, and separation performance. Experimental measurements and theoretical analyses have been used to quantify the FFV of polymers, but challenges remain because of their limitations. Experimental measurements are laborious and based on semi empirical equations, while Bondi’s group contribution theory involves ambiguities like the determination of van der Waals volume and the choice of factor values in the theoretical equation. To efficiently evaluate the FFV of polymers, this study utilizes high-throughput molecular dynamics (MD) simulations to build a large dataset regarding polymer’s FFV. Based on this large dataset, we further build machine learning (ML) models to establish the composition-structure relation. Inspired by group contribution theory which correlates polymer’s functional groups to FFV, our ML models correlate polymer’s substructures or physico-chemical indexes to FFV. Here, our study first benchmarks the MD simulation protocol to obtain reliable FFV of polymers and then carries out high-throughput MD simulations for more than 6,500 homopolymers and 1,400 polyamides. Such a large and diverse dataset makes the well-trained ML models more generalizable, compared with the group contribution theory. The efficiency of a feed forward neural network model is further demonstrated by applying it to a hypothetical polyimide dataset of more than 8 million chemical structures. The predicted FFVs of hypothetical polyimides are further validated by MD simulations. The obtained FFVs of the 8 million polymers, plus their previously reported gas separation performances, demonstrate the promising capability of ML virtual screening for the discovery of polymer membranes with exceptional permeability/selectivity.

36 MATERIALS SCIENCE↗

DOE OSTI2026

Latent Twins

Over the past decade, scientific machine learning has transformed the development of mathematical and computational frameworks for analyzing, modeling, and predicting complex systems. From inverse problems to numerical partial differential equations (PDEs), dynamical systems, and model reduction, these advances have pushed the boundaries of what can be simulated. Yet they have often progressed in parallel, with representation learning and algorithmic solution methods evolving largely as separate pipelines. With Latent Twins, we propose a unifying mathematical framework that creates a hidden surrogate in latent space for the underlying equations. Whereas digital twins mirror physical systems in the digital world, Latent Twins mirror mathematical systems in a learned latent space governed by operators. Through this lens, classical modeling, inversion, model reduction, and operator approximation all emerge as special cases of a single principle. We establish the fundamental approximation properties of Latent Twins for both ordinary differential equations (ODEs) and PDEs and demonstrate the framework across three representative settings: (i) canonical ODEs, capturing diverse dynamical regimes; (ii) a PDE benchmark using the shallow-water equations, contrasting Latent Twin simulations with deep operator network and forecasts with a four-dimensional variational method baseline; and (iii) a challenging real-data geopotential reanalysis dataset, reconstructing and forecasting from sparse, noisy observations. Latent Twins provide a compact, interpretable surrogate for solution operators that evaluate across arbitrary time gaps in a single-shot, while remaining compatible with scientific pipelines such as assimilation, control, and uncertainty quantification. Looking forward, this framework offers scalable, theory-grounded surrogates that bridge data-driven representation learning and classical scientific modeling across disciplines.

Latent Twins↗

DOE OSTI2024

Machine Learning-Based Identification of the Interface Regions for Coupling Local and Nonlocal Models

Local-nonlocal coupling approaches provide a means to combine the computational efficiency of local models and the accuracy of nonlocal models. However, the coupling process can be challenging, requiring expertise to identify the interface between local and nonlocal regions. Here, this study introduces a machine learning-based approach to automatically detect the regions in which the local and nonlocal models should be used. The method uses loading functions evaluated at grid points to decide the model selection at those points. Training of the networks is based on datasets provided by classes of loading functions for which reference coupling configurations are computed using accurate coupled solutions, where accuracy is measured in terms of the relative error between the solution to the coupling approach and the solution to the nonlocal model. We study two approaches that vary in data structure. The first, the full-domain input data approach, uses the entire load vector and outputs a complete label vector, performing a global classification. The second, a window-based approach, processes loads into windows and addresses the problem as a node-wise classification where each window's central point is classified individually. The classification problems are solved via deep learning algorithms based on convolutional neural networks. The performance of these approaches is studied on one-dimensional numerical examples using F1-scores and accuracy metrics. Notably, the windowing approach achieves an accuracy of 0.96 and an F1-score of 0.97, highlighting its potential to automate coupling processes effectively and enhance computational efficiency in material science applications.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Models and Processes to Extract Drug-like Molecules From Natural Language Text

Researchers worldwide are seeking to repurpose existing drugs or discover new drugs to counter the disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). A promising source of candidates for such studies is molecules that have been reported in the scientific literature to be drug-like in the context of viral research. However, this literature is too large for human review and features unusual vocabularies for which existing named entity recognition (NER) models are ineffective. We report here on a project that leverages both human and artificial intelligence to detect references to such molecules in free text. We present 1) a iterative model-in-the-loop method that makes judicious use of scarce human expertise in generating training data for a NER model, and 2) the application and evaluation of this method to the problem of identifying drug-like molecules in the COVID-19 Open Research Dataset Challenge (CORD-19) corpus of 198,875 papers. We show that by repeatedly presenting human labelers only with samples for which an evolving NER model is uncertain, our human-machine hybrid pipeline requires only modest amounts of non-expert human labeling time (tens of hours to label 1778 samples) to generate an NER model with an F-1 score of 80.5%—on par with that of non-expert humans—and when applied to CORD’19, identifies 10,912 putative drug-like molecules. This enriched the computational screening team’s targets by 3,591 molecules, of which 18 ranked in the top 0.1% of all 6.6 million molecules screened for docking against the 3CLPro protein.

60 APPLIED LIFE SCIENCES↗