Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network embeddings”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Low-cost embedded optical sensing systems for distribution transformer monitoring

There is a mounting need for low-cost monitoring with online sensing technologies to maintain grid reliability and uptime. Here we introduce an innovative low-cost, embedded optical sensing technology initially focused on transformers that was developed and demonstrated at a major electric utility, Con Edison. A version of it that can be retrofitted onto existing transformers in the field was also developed. Two new 500 kVA distribution network transformers were built with embedded fiber-optic (FO) sensors and qualified per industry standards. Vibration, temperature, and corrosion were key parameters monitored. The first transformer with embedded FO sensors was installed at a Con Edison facility and monitored at our team’s office. The second one was installed in an urban street-side underground location with online data processing/feature extraction algorithms and monitored through a wireless router. Additionally, an older transformer was also retrofitted. Data analysis was done on these transformers showing promising correlations with their corresponding loading cycles. Furthermore, key events such as the transformer primary-side energizing, and other events were detected. In general, the technology was demonstrated over 6 months across the 3 transformers instrumented with promising results. Thus, it has the potential to enable predictive maintenance for transformers and other grid assets.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2022

Bimetallic NiCo boride nanoparticles confined in a MXene network enable efficient ambient ammonia electrosynthesis

Ambient electrocatalytic nitrogen fixation is an emerging technology for green ammonia synthesis, but the absence of optimized, stable and performant catalysts can render its practical application challenging. Herein, bimetallic NiCo boride nanoparticles confined in MXene are shown to accomplish high-performance nitrogen reduction electrolysis. Taking advantage of the synergistic effect in specific compositions with unique electronic d and p orbits and typical architecture of rich nanosized particles embedded in the interconnected conductive network, the synthesized MXene@NiCoB composite demonstrates extensive improvements in nitrogen molecule chemisorption, active area exposure and charge transport. As a result, optimal NH 3 yield rate of 38.7 μg h -1 mgcat. -1 and Faradaic efficiency of 6.92% are acquired in 0.1 M Na 2 SO 4 electrolyte. Moreover, the great catalytic performance can be almost entirely maintained in the cases of repeatedly-cycled and long-term electrolysis. Theoretical investigations reveal that the nitrogen reduction reaction on MXene@NiCoB catalyst proceeds according to the distal pathway, with a distinctly-reduced energy barrier relative to the Co 2 B counterpart. In conclusion, this work may inspire a new route towards the rational catalyst design for the nitrogen reduction reaction.

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

Learning nuclear cross sections across the chart of nuclides with graph neural networks

We explore the use of deep learning techniques to learn how nuclear cross sections change as we add or remove protons and neutrons. As a proof of principle, we focus on the neutron-induced reactions in the fast energy regime. Our approach follows a two-stage learning framework. First, we apply representation learning to encode cross section data into a latent space using either variational autoencoders (VAEs) or implicit neural representations (INRs). Then, we train graph neural networks (GNNs) on the resulting embeddings to predict missing values across the nuclear chart by leveraging the topological structure of neighboring isotopes. We demonstrate accurate cross section predictions within a 9 × 9 block of missing nuclei. We also find that the optimal GNN training strategy depends on the type of latent representation used, with VAE embeddings performing best under end-to-end optimization in the original space, while INR embeddings achieve better results when the GNN is trained only in the latent space. Furthermore, using clustering algorithms, we map groups of latent vectors into regions of the nuclear chart and show that VAEs and INRs can discover some of the neutron magic numbers. These findings suggest that deep-learning models based on the representation encoding of cross sections combined with graph neural networks hold significant potential in augmenting nuclear theory models, e.g., by providing reliable estimates of covariances of cross sections, including cross-material covariances.

Machine learning↗
DOE OSTI2024

Estimating Interplanetary Magnetic Field Conditions at Mercury's Orbit From MESSENGER Magnetosheath Observations Using a Feedforward Neural Network

Abstract Mercury's small magnetosphere is embedded in the dynamic and intense solar wind environment characteristic of the inner heliosphere. Both the magnitude and orientation of the interplanetary magnetic field (IMF) significantly influence the solar wind‐magnetospheric interaction at Mercury, driving phenomena such as magnetic reconnection. The MErcury Surface, Space Environment, Geochemistry and Ranging (MESSENGER) spacecraft provided in‐situ magnetic field measurements of the solar wind, the magnetosheath, and the magnetosphere along each orbit. However, it is a challenge to directly assess the IMF's impact on Mercury's plasma environment due to the temporal separation between observations within the solar wind and the magnetosphere, especially in the absence of an upstream monitor. Here, we present a feedforward neural network (FNN) trained on a subset of magnetosheath observations to estimate the strength and orientation of the IMF upstream of the bow shock. Utilizing magnetosheath magnetic field, cylindrical spatial coordinates, and heliocentric distance measurements, the FNN predicts upstream IMF conditions with an score of 0.70 and mean averaged error of 5.3 nT, thereby greatly decreasing the temporal separation between IMF estimates and magnetospheric measurements throughout the MESSENGER mission. This approach yields IMF estimates for all magnetosheath data measured by MESSENGER, providing a useful tool for future investigations of the IMF impact on Mercury's magnetosphere. This method will be integrable with the dual‐spacecraft BepiColombo magnetosheath measurements, providing useful estimates of upstream IMF conditions particularly during the extended periods in which neither spacecraft sample the solar wind. Our results demonstrate the utility of machine learning techniques on advancing space science research.

Bowers, Charles F.↗
DOE OSTI2023

Soft, malleable double diamond twin

A twin boundary (TB) is a common low energy planar defect in crystals including those with the atomic diamond structure (C, Si, Ge, etc.). We study twins in a self-assembled soft matter block copolymer (BCP) supramolecular crystal having the double diamond (DD) structure, consisting of two translationally shifted, interpenetrating diamond networks of the minority polydimethyl siloxane block embedded in a polystyrene block matrix. The coherent, low energy, mirror-symmetric double tubular network twin has one minority block network with its nodes offset from the (222) TB plane, while nodes of the second network lie in the plane of the boundary. The offset network, although at a scale about a factor of 10 3 larger, has precisely the same geometry and symmetry as a (111) twin in atomic single diamond where the tetrahedral units spanning the TB retain nearly the same strut (bond) lengths and strut (bond) angles as in the normal unit cell. In DD, the second network undergoes a dramatic restructuring—the tetrahedral nodes transform into two new types of mirror-symmetric nodes (pentahedral and trihedral) which alternate and link to form a hexagonal mesh in the plane of the TB. The collective reorganization of the supramolecular packing highlights the hierarchical structure of ordered BCP phases and emphasizes the remarkable malleability of soft matter.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Multi-View Convolutional Neural Network for Data Spoofing Cyber-Attack Detection in Distribution Synchrophasors

Security of Distribution Synchrophasors Data (DSD) is of paramount importance as the data is used for critical smart grid applications including situational awareness, advanced protection, and dynamic control. Unfortunately, the DSD are attractive targets for malicious attackers aiming to damage grid. Data spoofing is a new class of deceiving attack, where the DSD of one Phasor Measurement Units (PMUs) is tampered by other PMUs thereby spoiling measurement based applications. In order to address this issue, a source authentication based data spoofing attack detection method is proposed using Multi-view Convolutional Neural Network (MCNN). First, common components embedded in raw frequency measurements from DSD are removed by Savitzky-Golay (SG) filter. Second, fast S transform (FST) is utilized to extract representative spatial fingerprints via time frequency analysis. Third, the spatial fingerprint is fed to MCNN, which combines dilated and standard convolutions for automatic feather extraction and source identification. Finally, according to the output of MCNN, spoofing attack detection is performed via threshold criterion. Extensive experiments with actual DSD from multiple locations in FNET/Grideye are conducted to verify the effectiveness of the proposed method.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

The LSST AGN Data Challenge: Selection Methods

Abstract Development of the Rubin Observatory Legacy Survey of Space and Time (LSST) includes a series of Data Challenges (DCs) arranged by various LSST Scientific Collaborations that are taking place during the project's preoperational phase. The AGN Science Collaboration Data Challenge (AGNSC-DC) is a partial prototype of the expected LSST data on active galactic nuclei (AGNs), aimed at validating machine learning approaches for AGN selection and characterization in large surveys like LSST. The AGNSC-DC took place in 2021, focusing on accuracy, robustness, and scalability. The training and the blinded data sets were constructed to mimic the future LSST release catalogs using the data from the Sloan Digital Sky Survey Stripe 82 region and the XMM-Newton Large Scale Structure Survey region. Data features were divided into astrometry, photometry, color, morphology, redshift, and class label with the addition of variability features and images. We present the results of four submitted solutions to DCs using both classical and machine learning methods. We systematically test the performance of supervised models (support vector machine, random forest, extreme gradient boosting, artificial neural network, convolutional neural network) and unsupervised ones (deep embedding clustering) when applied to the problem of classifying/clustering sources as stars, galaxies, or AGNs. We obtained classification accuracy of 97.5% for supervised models and clustering accuracy of 96.0% for unsupervised ones and 95.0% with a classic approach for a blinded data set. We find that variability features significantly improve the accuracy of the trained models, and correlation analysis among different bands enables a fast and inexpensive first-order selection of quasar candidates.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2026

Actinium–DOTA coordination in water from hybrid ML/MM: Structure, free energies, and water-exchange pathways

Quantitative simulation of trivalent ƒ-block chelates in water remains challenging because bonded and non-bonded force-field models make different approximations for coordination structure, exchange dynamics, and ion–ligand interactions in highly charged systems. Here, we develop a hybrid machine-learning/molecular-mechanics (ML/MM) framework for Ac 3+ –DOTA in explicit solvent by training an E(3)-equivariant neural network potential (MACELES) on mechanically embedded QM/MM data for Ac aquo and Ac–DOTA species and coupling it to NAMD 2.14 with particle-mesh Ewald electrostatics. Nanosecond ML/MM trajectories remain numerically stable and preserve chelate integrity, yielding a compact DOTA inner shell with an inner-sphere water coordination number of CN Ac,O w ≈ 1.7 arising from a dynamic equilibrium between one- and two-water states (37.5% and 59.9% of frames; three waters 2.5%). A 5 ns potential of mean force shows two low-lying basins at CN Ac,O w ≈ 1 and CN Ac,O w ≈ 2. DFT end-state free energies are consistent with the ML/MM profile, and DFT minimum-energy paths provide a qualitative electronic-structure reference for the observed basin connectivity. State-resolved kinetics reveal picosecond water-exchange pathways that couple hydration changes to transient DOTA arm fluctuations, and training-set comparisons show that temperature-matched Ac–DOTA data optimize energy/force accuracy while more diverse solvated data improve charge prediction. Overall, the present hybrid ML/MM model provides a practical description of Ac 3+ –DOTA hydration thermodynamics and short-time exchange behavior in explicit water at MD-like cost.

Actinium↗
DOE OSTI2024

A Self-Healing, Flowable, Yet Solid Electrolyte Suppresses Li-Metal Morphological Instabilities

In this article, lithium metal (Li 0 ) solid-state batteries encounter implementation challenges due to dendrite formation, side reactions, and movement of the electrode–electrolyte interface in cycling. Notably, voids and cracks formed during battery fabrication/operation are hot spots for failure. Here, a self-healing, flowable yet solid electrolyte composed of mobile ceramic crystals embedded in a reconfigurable polymer network is reported. This electrolyte can auto-repair voids and cracks through a two-step self-healing process that occurs at a fast rate of 5.6 µm h -1 . A dynamical phase diagram is generated, showing the material can switch between liquid and solid forms in response to external strain rates. The flowability of the electrolyte allows it to accommodate the electrode volume change during Li 0 stripping. Simultaneously, the electrolyte maintains a solid form with high tensile strength (0.28 MPa), facilitating the regulation of mossy Li 0 deposition. The chemistries and kinetics are studied by operando synchrotron X-ray and in situ transmission electron microscopy (TEM). Solid-state NMR reveals a dual-phase ion conduction pathway and rapid Li + diffusion through the stable polymer-ceramic interphase. This designed electrolyte exhibits extended cycling life in Li 0 –Li 0 cells, reaching 12 000 h at 0.2 mA cm -2 and 5000 h at 0.5 mA cm -2 . Furthermore, owing to its high critical current density of 9 mA cm -2 , the Li 0 –LiNi 0.8 Mn 0.1 Co 0.1 O 2 (NMC811) full cell demonstrates stable cycling at 5 mA cm -2 for 1100 cycles, retaining 88% of its capacity, even under near-zero stack pressure conditions.

25 ENERGY STORAGE↗
DOE OSTI2022

Scalable algorithms for physics-informed neural and graph networks

Physics-informed machine learning (PIML) has emerged as a promising new approach for simulating complex physical and biological systems that are governed by complex multiscale processes for which some data are also available. In some instances, the objective is to discover part of the hidden physics from the available data, and PIML has been shown to be particularly effective for such problems for which conventional methods may fail. Unlike commercial machine learning where training of deep neural networks requires big data, in PIML big data are not available. Instead, we can train such networks from additional information obtained by employing the physical laws and evaluating them at random points in the space–time domain. Such PIML integrates multimodality and multifidelity data with mathematical models, and implements them using neural networks or graph networks. Here, we review some of the prevailing trends in embedding physics into machine learning, using physics-informed neural networks (PINNs) based primarily on feed-forward neural networks and automatic differentiation. For more complex systems or systems of systems and unstructured data, graph neural networks (GNNs) present some distinct advantages, and here we review how physics-informed learning can be accomplished with GNNs based on graph exterior calculus to construct differential operators; we refer to these architectures as physics-informed graph networks (PIGNs). We present representative examples for both forward and inverse problems and discuss what advances are needed to scale up PINNs, PIGNs and more broadly GNNs for large-scale engineering problems.

42 ENGINEERING↗
DOE OSTI2024

E(n)-Equivariant cartesian tensor message passing interatomic potential

Machine learning potential (MLP) has been a popular topic in recent years for its capability to replace expensive first-principles calculations in some large systems. Meanwhile, message passing networks have gained significant attention due to their remarkable accuracy, and a wave of message passing networks based on Cartesian coordinates has emerged. However, the information of the node in these models is usually limited to scalars, and vectors. In this work, we propose High-order Tensor message Passing interatomic Potential (HotPP), an E(n) equivariant message passing neural network that extends the node embedding and message to an arbitrary order tensor. By performing some basic equivariant operations, high order tensors can be coupled very simply and thus the model can make direct predictions of high-order tensors such as dipole moments and polarizabilities without any modifications. The tests in several datasets show that HotPP not only achieves high accuracy in predicting target properties, but also successfully performs tasks such as calculating phonon spectra, infrared spectra, and Raman spectra, demonstrating its potential as a tool for future research.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Multimodal representation learning for predicting molecule–disease relations

Motivation: Predicting molecule–disease indications and side effects is important for drug development and pharmacovigilance. Comprehensively mining molecule–molecule, molecule–disease and disease–disease semantic dependencies can potentially improve prediction performance. Methods: We introduce a Multi-Modal REpresentation Mapping Approach to Predicting molecular-disease relations (M2REMAP) by incorporating clinical semantics learned from electronic health records (EHR) of 12.6 million patients. Specifically, M2REMAP first learns a multimodal molecule representation that synthesizes chemical property and clinical semantic information by mapping molecule chemicals via a deep neural network onto the clinical semantic embedding space shared by drugs, diseases and other common clinical concepts. To infer molecule–disease relations, M2REMAP combines multimodal molecule representation and disease semantic embedding to jointly infer indications and side effects. Results: We extensively evaluate M2REMAP on molecule indications, side effects and interactions. Results show that incorporating EHR embeddings improves performance significantly, for example, attaining an improvement over the baseline models by 23.6% in PRC-AUC on indications and 23.9% on side effects. Further, M2REMAP overcomes the limitation of existing methods and effectively predicts drugs for novel diseases and emerging pathogens. Availability and implementation: The code is available at https://github.com/celehs/M2REMAP, and prediction results are provided at https://shiny.parse-health.org/drugs-diseases-dev/.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Multi-state Catalysts Modulated by Mechanical Force (Final Report)

The development of more efficient catalytic processes and new approaches to control catalytic activity and selectivity are central to the realization of more selective, atom economic, and energy efficient routes to value added chemicals and polymers. The reactivity and selectivity of a transition metal catalyst is intimately related to the ligand-sphere geometry and, in many cases, the ideal ligand geometry for one step of a catalytic cycle is poorly matched to the ideal ligand geometry for another, resulting in sub-optimal efficiency. Macroscopic mechanical forces are both large, potentially much larger than interatomic forces, and are directional and localized to an extent that differentiates them from other forms of energy input such as heat or light. As such, mechanical force represents a heretofore untapped approach to modulate catalyst geometry, with the potential to reversibly modulate catalyst geometry on the timescale of catalytic turnover or monomer enchainment. This project has addressed the fundamental challenges in material-to-molecule strain coupling associated with the development of a new class of mechanically responsive catalysts (mechanocatalysts) in which active organotransition metal catalysts are strategically embedded in a flexible polymer network such that application of external mechanical force (stretching or deformation) leads to modulation of catalyst geometry, and hence reactivity and selectivity. Our efforts during the tenure of this grant were directed toward the elucidation of force-reactivity relationships of elementary transformations that occur within the first coordination sphere of a transition metal complex employing stiff stilbene photoswitches tethered to a flexible bidentate phosphine ligand derived from MeOBiphep as molecular force probes which provide a range of compressive and extension forces to the coupled transition metal complex depending on the geometry of the stiff stilbene and length of the tethering chains. During the tenure of this grant, we have quantified the rate of C(sp 2 )-C(sp 2 ) reductive elimination from platinum(II) diaryl complexes containing bis(phosphine) force probe ligands as a function of mechanical force; compressive forces decreased the rate of reductive elimination whereas extension forces increased the rate relative to the strain-free MeOBiphep complex with a 3.4-fold change in rate over a ~290 pN range of restoring forces. In a similar manner, we have quantified the rate of oxidative addition of bromobenzene to low-ligated palladium(0) complexes containing force probe ligands as a function of mechanical force; compressive forces increase the rate of oxidative addition, whereas tensile forces decrease the rate with a ~6 fold change in rate across ~340 pN of force applied to the complexes. In both cases, experimental and computational analyses argue strongly against any significant force-induced perturbation of ground state geometry within the first coordination sphere of the reactant complexes. Rather, the force/rate behavior observed for these transformations across these ranges of forces is attributed to the coupling of force to the nuclear motion comprising the reaction coordinates for reductive elimination and oxidative addition. These results together inform the development of catalysts whose activity can be tuned by an external force that is adjusted within a catalytic cycle and suggest opportunities to experimentally map geometry changes associated with reactions in transition metal complexes and potential strategies for force-modulated catalysis.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2020

Model-Agnostic Algorithm for Real-Time Attack Identification in Power Grid using Koopman Modes

Malicious activities on measurements from sensors like Phasor Measurement Units (PMUs) can mislead the control center operator into taking wrong control actions resulting in disruption of operation, financial losses, and equipment damage. In particular, false data attacks initiated during power systems transients caused due to abrupt changes in load and generation can fool the conventional model-based detection methods relying on thresholds comparison to trigger an anomaly. In this paper, we propose a Koopman mode decomposition (KMD) based algorithm to detect and identify false data attacks in real-time. The Koopman modes (KMs) are capable of capturing the nonlinear modes of oscillation in the transient dynamics of the power networks and reveal the spatial embedding of both natural and anomalous modes of oscillations in the sensor measurements. The Koopman-based spatio-temporal nonlinear modal analysis is used to filter out the false data injected by an attacker. The performance of the algorithm is illustrated on the IEEE 68-bus test system using synthetic attack scenarios generated on GridSTAGE, a recently developed multivariate spatio-temporal data generation framework for simulation of adversarial scenarios in cyber-physical power systems.

Nandanoori, Sai Pushpak↗
DOE OSTI2021

Automating Discovery of Physics-Informed Neural State Space Models via Learning and Evolution

Recent works exploring deep learning application to dynamical systems modeling have demonstrated that embedding physical priors into neural networks can yield more effective, physically-realistic, and data-efficient models. However, in the absence of complete prior knowledge of a dynamical system's physical characteristics, determining the optimal structure and optimization strategy for these models can be difficult. In this work, we explore methods for discovering neural state space dynamics models for system identification. Starting with a design space of block-oriented state space models and structured linear maps with strong physical priors, we encode these components into a model genome alongside network structure, penalty constraints, and optimization hyperparameters. Demonstrating the overall utility of the design space, we employ an asynchronous genetic search algorithm that alternates between model selection and optimization and obtains accurate physically consistent models of three physical systems: an aerodynamics body, a continuous stirred tank reactor, and a two tank interacting system.

genetic algorithms, neural architecture search, ne↗
DOE OSTI2023

Systems and methods for customizing kernel machines with deep neural networks

A method including receiving an input data set. The input data set can include one of a feature domain set or a kernel matrix. The method also can include constructing dense embeddings using: (i) Nyström approximations on the input data set when the input data set comprises the kernel matrix, and (ii) clustered Nyström approximations on the input data set when the input data set comprises the feature domain set. The method additionally can include performing representation learning on each of the dense embeddings using a multi-layer fully-connected network for each of the dense embeddings to generate latent representations corresponding to each of the dense embeddings. The method further can include applying a fusion layer to the latent representations corresponding to the dense embeddings to generate a combined representation. The method additionally can include performing classification on the combined representation. Other embodiments of related systems and methods are also disclosed.

Song, Huan↗
DOE OSTI2024

Equipping Neural Network Surrogates with Uncertainty for Propagation in Physical Systems

Coarse-grained or filtered models typically rely on closure models to account for unresolved scales. For instance, large eddy simulation for modeling turbulent fluid flows explicitly resolves the largest scales, but requires modeling closure terms to account for the sub-filter scales. With the vast amount of data available from high-fidelity simulations, there are unique opportunities to leverage data-driven modeling techniques to formulate expressive and flexible closure models. Despite their flexibility, data-driven models struggle in domain shift settings, i.e. when deployed in configurations not captured in the training dataset. In particular, the efficacy of neural network surrogates is difficult to assess a priori due to the deterministic, point-estimate nature of predictions. In high-consequence applications, such models require reliable uncertainty estimates in the data-informed and out-of-distribution regimes. To quantify uncertainties in both regimes, we employ Bayesian neural networks which are able to capture both epistemic and aleatoric uncertainties. We will discuss challenges associated with the training and evaluation of these networks. Furthermore, we will discuss uncertainty embedding strategies to enable efficient sampling and propagation of uncertainty through high-fidelity simulations.

Bayesian neural networks↗