Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Machine Learning Models”

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 1,099 records · Page 61
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

AGFormer: Adaptive Spatiotemporal graph informed transformer for multi-reservoir inflow forecasting

Accurate reservoir inflow forecasting is crucial for effective water resource management, yet most machine learning models focus on single-reservoir prediction and overlook spatial dependencies among hydrologically connected reservoirs. Here, we propose AGFormer (Adaptive Graph-Informed Transformer), an end-to-end framework that integrates adaptive graph learning with temporal sequence modeling for multi-reservoir inflow forecasting. A shared encoder and graph attention mechanism generate reservoir-specific embeddings, which are then processed by the Transformer-based encoder–decoder for multi-step inflow forecasting. We also introduce a pretraining paradigm to learn robust temporal embeddings from misaligned historical records. Evaluated on 30 reservoirs in the Upper Colorado River Basin, AGFormer achieves superior seven-day-ahead forecasts, with NSE > 0.75 for 20 reservoirs—outperforming Encoder–Decoder LSTM, GCN+LSTM, and Transformer baselines. Adaptive graph learning captures dynamic inter-reservoir dependencies, and feature attribution aligns with snowmelt-driven hydrology. Incorporating forecasted meteorological inputs further enhances accuracy, demonstrating AGFormer’s potential to support reservoir management under dynamic hydrological conditions.

Adaptive graph learning↗
DOE OSTI2026

Roadmap and Benchmarking: Privacy in Federated Load Forecasting

Data-driven techniques for energy demand forecasting continue to emerge with promising impacts on distribution grid planning. However, the development of robust and generalizable machine learning models requires that representative high quality training data are available. Distributed energy resources have begun to embed intelligence, gathering large amounts of data on customer demand, behavior, and household devices that are connected to the grid. Though utilities aggregate meter-level demand data for load shaping, demand response, outage management, reliability planning, and billing applications, there lies an inherent privacy concern in sharing consumption data that may identify individual consumer behavioral patterns. Hence, while sharing the data is crucial, the private sensitive customer data must be safeguarded from being exposed or manipulated. In this study, we propose a roadmap for implementing a based privacy preserving framework to support the advancement of data-driven analytics in data-sensitive distributed energy resources environments. The roadmap incorporates federated learning–a distributed training framework, differential privacy–a statistical framework that provides guarantees to safeguard the leakage of sensitive data, secure multiparty computation and homomorphic encryption– techniques for encrypting model gradients and applying secure aggregation on the server. Moreover, we perform baseline experiments on the federated short-term load forecasting (STLF) task using open-source residential load profile datasets, offering insights into the challenges of integrating differential privacy into federated learning.

Abebe, Waqwoya [Oak Ridge National Laboratory (ORN↗
DOE OSTI2022

Machine learning magnetism classifiers from atomic coordinates

The determination of magnetic structure poses a long-standing challenge in condensed matter physics and materials science. Experimental techniques such as neutron diffraction are resource-limited and require complex structure refinement protocols, while computational approaches such as first-principles density functional theory (DFT) need additional semi-empirical correction, and reliable prediction is still largely limited to collinear magnetism. Here, we present a machine learning model that aims to classify the magnetic structure by inputting atomic coordinates containing transition metal and rare earth elements. By building a Euclidean equivariant neural network that preserves the crystallographic symmetry, the magnetic structure (ferromagnetic, antiferromagnetic, and nonmagnetic) and magnetic propagation vector (zero or non-zero) can be predicted with an average accuracy of 77.8% and 73.6%. In particular, a 91% accuracy is reached when predicting no magnetic ordering even if the structure contains magneticelement(s). Ourworkrepresents onestepforwardtosolvingthegrand challenge of full magnetic structure determination.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2021

Machine learning reveals sequence-function relationships in family 7 glycoside hydrolases

Family 7 glycoside hydrolases (GH7) are among the principal enzymes for cellulose degradation in nature and industrially. These important enzymes are often bimodular, including a catalytic domain and carbohydrate binding module (CBM) attached via a flexible linker, and exhibit an active site that binds cello-oligomers of up to ten glucosyl moieties. GH7 cellulases consist of two major subtypes: cellobiohydrolases (CBH) and endoglucanases (EG). Despite the critical importance of GH7 enzymes, there remain gaps in our understanding of how GH7 sequence and structure relate to function. Here, we employed machine learning to gain data-driven insights into relationships between sequence, structure, and function across the GH7 family. Machine-learning models, trained only on the number of residues in the active-site loops as features, were able discriminate GH7 CBHs and EGs with up to 99% accuracy, demonstrating that the lengths of loops A4, B2, B3, and B4 strongly correlate with functional subtype across the GH7 family. Classification rules were derived such that specific residues at 42 different sequence positions each predicted the functional subtype with accuracies surpassing 87%. A random forest model trained on residues at 19 positions in the catalytic domain predicted the presence of a CBM with 89.5% accuracy. Our ML results recapitulate, as top-performing features, a substantial number of the sequence positions determined by previous experimental studies to play vital roles in GH7 activity. We surmise that the yet-to-be-explored sequence positions among the top-performing features also contribute to GH7 functional variation and may be exploited to understand and manipulate function.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Class imbalance in out-of-distribution datasets: Improving the robustness of the TextCNN for the classification of rare cancer types

In the last decade, the widespread adoption of electronic health record documentation has created huge opportunities for information mining. Natural language processing (NLP) techniques using machine and deep learning are becoming increasingly widespread for information extraction tasks from unstructured clinical notes. Disparities in performance when deploying machine learning models in the real world have recently received considerable attention. In the clinical NLP domain, the robustness of convolutional neural networks (CNNs) for classifying cancer pathology reports under natural distribution shifts remains understudied. In this research, we aim to quantify and improve the performance of the CNN for text classification on out-of-distribution (OOD) datasets resulting from the natural evolution of clinical text in pathology reports. We identified class imbalance due to different prevalence of cancer types as one of the sources of performance drop and analyzed the impact of previous methods for addressing class imbalance when deploying models in real-world domains. Our results show that our novel class-specialized ensemble technique outperforms other methods for the classification of rare cancer types in terms of macro F1 scores. We also found that traditional ensemble methods perform better in top classes, leading to higher micro F1 scores. Based on our findings, we formulate a series of recommendations for other ML practitioners on how to build robust models with extremely imbalanced datasets in biomedical NLP applications.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

A high-throughput approach for statistical process optimization in Laser Powder Bed Fusion

Process variability is inherent in metal additive manufacturing (AM). However, it is often overlooked in process optimization frameworks, constraining the understanding of process uncertainties and their influence on parameter selection. To address this, we present an integrated framework that combines high-throughput single-track experiments, GAN-based melt pool geometry extraction, robust statistical and machine learning modeling, and uncertainty-quantified process mapping. Process variability is characterized through single-track melt pool behaviors, and its influence on defect formation is systematically quantified to enable statistically guided process parameter optimization. This approach is demonstrated on Laser Powder Bed Fusion (L-PBF) of stainless steel 316L, effectively capturing the interplay between process parameters, melt pool variability, and defect probability. By integrating uncertainty quantification into process optimization, this study provides a structured methodology for addressing variability challenges in AM quality control, ultimately contributing to enhanced manufacturing reliability.

Laser Powder Bed Fusion↗
DOE OSTI2024

Enriching OpenStreetMap network data for transportation applications: Insights into the impact of urban congestion on accessibility

OpenStreetMap (OSM) data is a valuable open-source resource for various transportation, traffic, and planning applications. However, OSM network data lack operating traffic speed information, which is critical for transport planning and operations. Addressing this shortcoming, this study leverages commercial vendor data (to serve as ground truth) with exogenous, open-source variables characterizing local transport infrastructure, land use, and demographic information to predict average congested traffic speeds on OSM networks. Three machine-learning models were tested and estimated for OSM links with and without speed limit information in the Denver metropolitan region. Among these, XGBoost performed best, with mean absolute errors of 3.27 and 3.62 mph for links with and without speed limits, respectively. The developed models accurately predicted traffic speeds for different hours and days of the week compared to ground truth data. Using these predicted speeds, drive accessibility scores were computed for the Denver region for different time periods using the Mobility Energy Productivity (MEP) metric to understand the impact of congestion on energy-efficient accessibility. Results show that congestion-adjusted drive accessibility can be significantly lower compared to accessibility calculated using free flow speeds. Specifically, weekday evening hours saw a 42 % drop in accessibility due to reduced speeds, particularly around downtown Denver. Across the Denver metro region, approximately half as many opportunities and jobs are accessible in under 20 min by car during the evening peak period relative to free flow conditions. These findings underscore the importance of using congestion-adjusted operating speeds rather than speed limits in accessibility calculations, as reliance on speed limits can substantially overestimate energy-efficient drive accessibility in large, car-centric cities susceptible to significant congestion. In conclusion, the methodology presented here could further enrich OSM network data, making them useful for an even broader range of transportation applications.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2024

Modeling plasticity-mediated void growth at the single crystal scale: A physics-informed machine learning approach

Modeling the evolution of voids during plastic flow as well as their effects on plastic dissipation is critical for both component manufacturing and lifetime estimation purposes. To this end, we propose a rate-dependent constitutive model to homogenize the effects of semi-randomly distributed voids on single crystal plasticity whilst capturing void interaction and plastic anisotropy. Here, this present work focuses on the case of face centered cubic crystals to introduce an anisotropic gauge function applicable within the crystal plasticity formalism. The approach combines analytical methods to describe the micromechanics of the system in combination with symbolic regression to capture analytically intractable mechanisms from data. The hybrid framework uses a physics-informed genetic programming-based symbolic regression algorithm to solve a multiform optimization problem simultaneously producing a new gauge function and a new strain rate equation. This is also a multi-objective optimization problem with many competing objectives. A new search and selection step is introduced to the genetic algorithm that promotes convergence toward a global solution that better satisfies all the objectives. Overall, the symbolic equations produced leverage data-driven methods to achieve greater accuracy than comparable alternatives on an analytically intractable problem while maintaining model transparency.

36 MATERIALS SCIENCE↗
DOE OSTI2021

High-throughput screening of multifunctional nanocoatings based on combinations of polyphenols and catecholamines

Biomimetic surface coatings based on plant polyphenols and catecholamines have been used broadly in a variety of applications. However, the lack of a rational cost-effective platform for screening these coatings and their properties limits the true potential of these functional materials to be unleashed. Here, we investigated the oxidation behavior and coating formation ability of a library consisting of 45 phenolic compounds and catecholamines. UV–vis spectroscopy demonstrated significant acceleration of oxidation and polymerization under UV irradiation. We discovered that several binary mixtures resulted in non-additive behavior (synergistic or antagonistic effect) yielding much thicker or thinner coatings than individual compounds measured by ellipsometry. To investigate the properties of coatings derived from new combinations, we used a miniaturized high-throughput strategy to screen 2,532 spots coated with single, binary, and ternary combinations of coating precursors in one run. We evaluated the use of machine learning models to learn the relation between the chemical structure of the precursors and the thickness of the nanocoatings. Formation and stability of nanocoatings were investigated in a high-throughput manner via discontinuous dewetting. 30 stable combinations (hits) were used to tune the surface wettability and to form water droplet microarray and spot size gradients of water droplets on the coated surface. No toxicity was observed against eukaryotic HeLa cells and Pseudomonas aeruginosa (strain PA30) bacteria after 24 h incubation at 37 C. The strategy introduced here for high-throughput screening of nanocoatings derived from combinations of coating precursors enables the discovery of new functional materials for various applications in science and technology in a cost-effective miniaturized manner.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Unsupervised domain adaptation for radioisotope identification in gamma spectroscopy

Training machine learning models for radioisotope identification using gamma spectroscopy remains an elusive challenge for many practical applications, largely stemming from the difficulty of acquiring and labeling large, diverse experimental datasets. Simulations can mitigate this challenge, but the accuracy of models trained on simulated data can deteriorate substantially when deployed to an out-of-distribution operational environment. In this study, we demonstrate that unsupervised domain adaptation (UDA) can improve the ability of a model trained on synthetic data to generalize to a new testing domain, provided unlabeled data from the target domain are available. Conventional supervised techniques are unable to utilize this data because the absence of isotope labels precludes defining a supervised classification loss. Instead, we first pretrain a spectral classifier using labeled synthetic data and subsequently leverage unlabeled target data to align the learned feature representations between the source and target domains. We compare a range of different UDA techniques, finding that minimizing the maximum mean discrepancy (MMD) between source and target feature vectors yields the most consistent improvement to testing scores. For instance, using a custom transformer-based neural network, we achieved a testing accuracy of $0.904 \pm 0.022$ on an experimental LaBr test set after performing unsupervised feature alignment via MMD minimization, compared to $0.754 \pm 0.014$ before alignment. Overall, our results highlight the potential of using UDA to adapt a radioisotope classifier trained on synthetic data for real-world deployment.

Lalor, Peter W.↗
DOE OSTI2025

Developing a digital twin framework for remotely monitoring nuclear reactor facilities

A digital twin must seek to represent all applicable functional components of the system of interest. Different expertise is required for understanding the physical system being modeled than the skills needed for transforming those models into a functional digital twin through physics modeling, machine learning analysis, and visualization. The diversity of knowledge requires a multi-disciplinary team to ensure all system details are captured. Team members also need a method to verify that the data they generate within their domain can be effectively communicated to professionals in other fields. To address this challenge, this work provides an approach for developing a digital twin framework to remotely monitoring nuclear facilities. Through this, general knowledge of the framework is presented along with two examples to solidify the process. The AGN-201 digital twin and microreactor digital twins provide varying levels of complexity in a potential nuclear facility, where common threads are identified and lessons learned are provided. The goal of this research is to aid future researchers by providing a formula for a successful digital twin and in turn reducing the development time of nuclear system digital twins, specifically for remote monitoring.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2022

Explainable machine learning of the underlying physics of high-energy particle collisions

We present an implementation of an explainable and physics-aware machine learning model capable of inferring the underlying physics of high-energy particle collisions using the information encoded in the energy-momentum four-vectors of the final state particles. We demonstrate the proof-of-concept of our White Box AI approach using a Generative Adversarial Network (GAN) which learns from a DGLAP-based parton shower Monte Carlo event generator. The constrained generator network architecture mimics the structure of a parton shower exhibiting similarities with Recurrent Neural Networks (RNNs). We show, for the first time, that our approach leads to a network that is able to learn not only the final distribution of particles, but also the underlying parton branching mechanism, i.e. the Altarelli-Parisi splitting function, the ordering variable of the shower, and the scaling behavior. While the current work is focused on perturbative physics of the parton shower, we foresee a broad range of applications of our framework to areas that are currently difficult to address from first principles in QCD. Examples include nonperturbative and collective effects, factorization breaking and the modification of the parton shower in heavy-ion, and electron-nucleus collisions.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2025

deadtrees.earth — An open-access and interactive database for centimeter-scale aerial imagery to uncover global tree mortality dynamics

Excessive tree mortality is a global concern and remains poorly understood as it is a complex phenomenon. We lack global and temporally continuous coverage on tree mortality data. Ground-based observations on tree mortality, e.g., derived from national inventories, are very sparse, and may not be standardized or spatially explicit. Earth observation data, combined with supervised machine learning, offer a promising approach to map overstory tree mortality in a consistent manner over space and time. However, global-scale machine learning requires broad training data covering a wide range of environmental settings and forest types. Low altitude observation platforms (e.g., drones or airplanes) provide a cost-effective source of training data by capturing high-resolution orthophotos of overstory tree mortality events at centimeter-scale resolution. Here, we introduce deadtrees.earth, an open-access platform hosting more than two thousand centimeter-resolution orthophotos, covering more than 1,000,000 ha, of which more than 58,000 ha are manually annotated with live/dead tree classifications. This community-sourced and rigorously curated dataset can serve as a comprehensive reference dataset to uncover tree mortality patterns from local to global scales using space-based Earth observation data and machine learning models. This will provide the basis to attribute tree mortality patterns to environmental changes or project tree mortality dynamics to the future. The open nature of deadtrees.earth, together with its curation of high-quality, spatially representative, and ecologically diverse data will continuously increase our capacity to uncover and understand tree mortality dynamics.

Citizen science↗
DOE OSTI2025

Porous carbon from lignocellulosic biomass with emphasis on corn plant waste residue for energy storage

The rising global demand for sustainable energy storage materials has driven the search for environmentally friendly and cost-effective electrode options. Hydrothermal conversion of lignocellulosic biomass has gained attention due to its low energy requirements and operation at relatively low temperatures, presenting a green alternative to traditional thermochemical methods. The resulting solid product, hydrochar, has been used as an adsorbent and soil amendment; however, chemical/thermal treatment significantly enhances its physical properties. These structural modifications transform hydrochar into an effective porous carbon electrode, offering abundant sites for electrolyte ion transport, critical for high-performance devices like supercapacitors and batteries. This review first discusses various waste biomass and sustainable feedstocks available globally. It compares two primary thermochemical conversion techniques, pyrolysis and hydrothermal carbonization/liquefaction, and examines their respective solid products, biochar and hydrochar, analyzing differences in their physical and chemical characteristics. The focus is placed on hydrochar, summarizing activation methods to produce porous carbon suitable for energy storage applications. Additionally, this review will include a dedicated section on the application of porous carbon derived from corn plant waste residue, considering that corn is one of the most abundant crops grown worldwide, which makes it an important and promising source for sustainable porous carbon production. The role of machine learning models in optimizing hydrothermal processes to produce high-quality hydrochar is also discussed, emphasizing how data-driven approaches can streamline process development. Finally, the review identifies the current challenges and prospects for lignocellulosic biomass-derived porous carbon as a sustainable electrode material in next-generation energy storage technologies.

25 ENERGY STORAGE↗
DOE OSTI2022

Automated defect identification in electroluminescence images of solar modules

Solar photovoltaic (PV) modules are susceptible to manufacturing defects, mishandling problems or extreme weather events that can limit energy production or cause early device failure. Trained professionals use electroluminescence (EL) images to identify defects in modules, however, field surveys or inline image acquisition can generate millions of EL images, which are infeasible to analyze by rote inspection. Here, we develop a rapid automatic computer vision pipeline (~0.5 seconds/module) to analyze EL images and identify defects including cracks, intra-cell defects, oxygen-induced defects, and solder disconnections. Defect identification is achieved with a machine learning model (Random Forest, ResNet models and YOLO) trained on 762 manually-labeled EL images of PV modules. We compare model performance on an imbalanced real-world validation set containing 134 EL images and determine that ResNet18 and YOLO are the optimal models; we next evaluated these models on a dedicated testing set (129 module images) with resulting macro F1 scores of 0.83 (ResNet18) and 0.78 (YOLO). Using a field EL survey of a PV power plant damaged in a vegetation fire, we analyze 18,954 EL images (2.4 million cells) and inspect the spatial distribution of defects on the solar modules. The results find increased frequency of ‘crack’, ‘solder’ and ‘intra-cell’ defects on the edges of the solar module closest to the ground after fire. We also find an abnormal increase of striation rings on cells which were assumed to be caused mainly in fabrication process. Our methods are published as open-source software. It can also be used to identify other kinds of defects or process different types of solar cells with minor modification on models by transfer learning.

14 SOLAR ENERGY↗
DOE OSTI2020

Reducing ridesourcing empty vehicle travel with future travel demand prediction

Ridesourcing services provide alternative mobility options in several cities. Their market share has grown exponentially due to the convenience they provide. The use of such services may be associated with car-light or car-free lifestyles. However, there are growing concerns regarding their impact on urban transportation operations performance due to empty, unproductive miles driven without a passenger (commonly referred to as deadheading). This paper is motivated by the potential to reduce deadhead mileage of ridesourcing trips by providing drivers with information on future ridesourcing trip demand. Future demand information enables the driver to wait in place for the next rider’s request without cruising around and contributing to congestion. A machine learning model is employed to predict hourly and 10-minute future interval travel demand for ridesourcing at a given location. Using future demand information, we propose algorithms to (i) assign drivers to act on received demand information by waiting in place for the next rider, and (ii) match these drivers with riders to minimize deadheading distance. Real-world data from ridesourcing providers in Austin, TX (RideAustin) and Chengdu, China (DiDi Chuxing) are leveraged. Results show that this process achieves 68%–82% and 53%–60% reduction of trip-level deadheading miles for the RideAustin and DiDi Chuxing sample operations respectively, under the assumption of unconstrained availability of short-term parking. Deadheading savings increase slightly as the maximum tolerable waiting time for the driver increases. Further, it is observed that significant deadhead savings per trip are possible, even when a small percent of the ridesourcing driver pool is provided with future ridesourcing demand information.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Discovery of Ternary Antimonides A–Al–Sb (A = Rb or Cs) with Desired Structural Motifs Guided by Machine Learning

Specific structural motifs in inorganic solids are often related to their targeted physical properties. For many classes of solids, such as Zintl phases and polar intermetallics, the crystal structures are diverse and not easy to predict. Various antimonides that are potential thermoelectric materials were proposed to be synthesizable on the basis of their estimated formation energies. Their structures were broadly classified as clathrate, channel, layered, or network through a machine learning model trained on existing ternary phases and features based on elemental properties using the sure independence screening and sparsifying operator algorithm. Through experimental validation, three new ternary antimonides were synthesized and confirmed to form layered structures: tetragonal RbAlSb 2 and CsAlSb 2 , which are isopointal but not isotypic to LiBSi 2 ; and monoclinic Rb 2 Al 2 Sb 3 , which adopts the Na 2 Al 2 Sb 3 -type structure. Finally, reinvestigation of the related compound Cs 2 In 2 Sb 3 revealed a low thermal conductivity and p-type semiconducting behavior.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Trust Not Verify? The Critical Need for Data Curation Standards in Materials Informatics

The importance of data curation has been recognized in multiple areas of research; however, the discussion of this important issue is only beginning to emerge in materials science. In this Perspective, we highlight the benefits of using the standardized data curation protocols in materials science and discuss current gaps in accurate and reproducible data reporting using case studies drawn from high-impact materials science papers and well-known databases such as the Crystallography Open Database (COD) and the Cambridge Structural Database (CSD). We argue that both experimental and computational materials scientists need to embrace a culture of rigorous data curation as part of modern research data management. We propose a sample data curation pipeline for materials chemistry and illustrate its use by creating two new materials chemistry databases. Here, we hope that this perspective will serve to catalyze further discussion and promote the continuous development of rigorous data curation practices within the materials science research community. We posit that adherence to best practices of data curation will promote and enhance the reliability, reproducibility, and integrity of materials research and enable the development of reliable AI and machine learning models that critically depend on the use of quality data.

Chemical structure↗