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.

SourceResearch connectionFrom yearThrough year

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

At least 397 records · Page 22

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Dual-Wavelength Simultaneous Patterning of Degradable Thermoset Supports for One-Pot Embedded 3D Printing

Vat photopolymerization (VP) techniques have enabled the fabrication of complex geometries while balancing high precision and fast processing times. 3D printed objects are traditionally built layer-by-layer with newly cured layers being structurally supported by previous ones. Fabricating unsupported features such as overhangs and arches risks misalignment and sagging, limiting the range of accessible designs. To overcome this issue, support structures are fabricated along with the primary object as temporary scaffolds that provide stability and conserve print fidelity. For VP specifically, patterning dissolvable sacrificial supports is attractive to avoid manual removal after printing. In this study, we demonstrate a base-degradable thermoset to pattern print supports in a one-pot formulation along with the primary structural material. Efficient printing is enabled using a dual-wavelength negative imaging (DWNI) DLP printer that patterns the degradable thermoset with visible light and the permanent network with UV light, which are simultaneously projected using a single digital micromirror device (DMD). Printed objects undergo thermal postprocessing to enhance the final conversion of the primary material, after which thermoset supports are degraded in a basic, aqueous solution. This approach provides a robust method for the dual-wavelength patterning of sacrificial thermoset supports, broadening the range of accessible 3D printable materials and geometries.

3D printing↗

DOE OSTI2020

Spectrum of embrittling potencies and relation to properties of symmetric-tilt grain boundaries

Grain boundaries in metallic materials can exist in a wide range of stable and metastable structures. In addition, the properties of a grain boundary may be altered through solute segregation. In this work, we present a formulation that combines the spectrum of embrittling potencies associated with solute segregation with site-occupancy statistics. As a prototype problem, we illustrate the relation between segregation and embrittlement in the case of S segregation to grain boundaries in Ni. To obtain a population of site segregation energies, we perform molecular statics calculations on 378 different symmetric-tilt grain boundaries and their free surface equivalents, using an embedded-atom method interatomic potential developed specifically for studying embrittlement. Our results show that it is important to consider both the energies associated with embrittlement and the probability of occupancy to describe the general embrittling nature of a grain boundary. When analyzed in isolation, certain grain boundaries show large embrittling potencies; however, that effect is diminished when the probability of S segregation to that grain boundary is considered within a polycrystal. We propose a new quantity, the embrittling estimator, which not only categorizes grain boundaries as embrittling or strengthening, but also considers site occupancy probabilities, so that the embrittlement behavior of grain boundaries within a network of grain boundaries can be compared. Finally, we examine the relationship between embrittlement behavior and innate grain boundary properties, such as the free volume, and find statistical evidence that the complex nature of embrittlement cannot be explained by linear correlations with excess volumes or energies. Ultimately, this combined approach provides a theoretical tool to assist grain boundary engineering of metastable alloys.

36 MATERIALS SCIENCE↗

DOE OSTI2020

Models and Strategies for Optimal Demand Side Management in the Chemical Industries

Deregulation and the increase of renewable electricity generation from wind and solar photovoltaics have transformed the U.S. electricity market. Economic and environmental benefits notwithstanding, the presence of renewables has increased variability and uncertainty on the supply side of the grid. Managing demand, rather than generation – a strategy referred to as “demand response (DR)” – is an attractive approach for mitigating this imbalance. DR efforts aim to reduce electricity usage during peak demand times, lessening stress on the grid. Industrial users are particularly attractive entities for DR participation since they present large, localized loads that can provide significant relief on grid demand and –unlike other large loads, such as buildings – are minimally dependent on human needs and preferences. In this project, we accomplished three main objectives. (1) We developed data-driven low-order DR scheduling-relevant dynamic models of chemical processes. Concurrently, we studied the formulation and solution of the associated optimal DR production scheduling problems. (a) A prototype air separation unit (ASU) model was used to generate simulated operating data for initial modeling efforts, which enabled the later use of industrial data for data-driven modeling. (b) We utilized Hammerstein-Wiener (HW) and Finite Step Response (FSR) models to represent nonlinear plant dynamics. (c) The HW models were linearized using exact linearization so they could potentially be embedded in power system models, which are formulated as mixed integer linear programs (MILPs). (d) We solved DR optimization problems under uncertainty and found that even naïve predictions of electricity price and product demand led to significant cost savings benefits. (2) Our DR scheduling optimization problem formulations are amenable to real-time solution. (a) We utilized Lagrangian Relaxation (LR) to efficiently solve the optimization problem by decoupling subproblems linked by complicating constraints. (b) We have achieved computation times for the 3-day DR scheduling problem of an ASU as low as 1.88 minutes. (3) Our representations of the DR behavior of chemical process as grid-level batteries were embedded in power system models. (a) For a small-scale grid, we found that incorporating the dynamics of the chemical plant in the optimal power flow calculations resulted in better resource management leading to up to 15% and 46% cost reduction for the grid and chemical plant operations, respectively, during periods of power line congestion. We have published several works dedicated to modeling and solving DR optimization problems from the user side. These were published in top peer-reviewed journals and are summarized in this report. The most recent work (and papers in preparation) considers DR scheduling from the grid side. Future efforts will consider networked plants (e.g., air separation units operating on a common pipeline) for DR participation, which is expected to amplify the capabilities of industrial DR participants to perform load-shifting. Our consideration of uncertainty in DR has inspired future directions in this area as well: we plan to develop multistage methods to fully account for the effects of uncertainty in DR scheduling.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2023

Decoding Zeolite Crystallization and Stage III in Nuclear Waste Glasses by Coupled Modeling and Experiments

Under specific conditions of pH and temperature, nuclear waste immobilization borosilicate glasses may exhibit a sudden acceleration in their corrosion kinetics (stage III)—a behavior that has been associated with the formation of zeolite crystals. Such accelerated dissolution may compromise the integrity of nuclear wasteforms placed in geological depositories. However, thus far, none of the available models is able to predict the thermodynamic propensity and kinetics of zeolite precipitation as a function of the solution conditions due to (i) a lack of fundamental knowledge regarding the nucleation & growth mechanisms of zeolitic phases, (ii) uncertainty regarding the compositions (types) of zeolites that may form and the rate-limiting step in their precipitation as a function of the solution conditions, and (iii) the complexities that arise due to the vast parametric space (i.e., solution chemistry, temperature, number of secondary phases, etc.) that encompass these systems under conditions of environmental exposure. To resolve these challenges, this project aimed to unambiguously identify the thermodynamic propensity for zeolite precipitation and the kinetics thereof as a function of the solution conditions (composition, pH, and temperature). To achieve this goal: 1) We identified the solution conditions and zeolite phases relevant to nuclear glass dissolution. 2) We performed a series of ab initio molecular dynamics (AIMD) simulations to compute the thermodynamic properties of a group of characteristic zeolites that features a large range of compositions, various hydration levels, a wide range of framework structures, and partial atomic site occupancies. 3) We released a first-of-a-kind self-consistent thermodynamic database that can be used to assess the kinetics and the stability fields of zeolitic phases within a Gibbs energy minimization (GEM) framework. 4) We developed a robust geochemical modeling method allowing us to predict the stability of secondary phases (including zeolites, calcium–silicate–hydrate gels, and clays) upon the dissolution of nuclear waste immobilization glasses. 5) We introduced a model that predicts the dissolution kinetics of a series of borosilicate nuclear waste immobilization glasses in terms of the topology of their atomic network. 6) We investigated the roles of the solution composition on the crystallization kinetics of phillipsite zeolites and tobermorite silicate hydrates. Via PNNL’s collaboration and engagement, this project directly supports DOE’s nuclear waste immobilization activities by offering a technical, science-based foundation that will (i) facilitate predictions of the long-term corrosion rates and extents of existing nuclear waste immobilization glasses to help ensure safe and successful vitrification operations, (ii) inform the development of advanced glass formulations with enhanced durability, and, (iii) enable cost-savings that result from making more decisive and hence less conservative predictions while offering higher levels of nuclear waste embedment in smaller, more compact glass volumes.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2023

HT Model Dataset

This website contains the dataset that was used for writing the manuscript "HT Model: Using the Molecular Transformer for predicting hydrotreating reactions" (PNNL-SA-186589) The dataset includes a collection of hydrotreating reactions compiled from 41 peer-reviewed literature sources. These sources contain experimental data related to hydrotreating reactions. These reactions involve the reaction of chemical compounds with hydrogen gas in the presence of a catalyst to remove heteroatoms or to convert specific functional groups. The dataset contains reactions both with and without reaction conditions. Reaction conditions refer to the specific parameters under which the reaction takes place, such as temperature and pressure. For each reaction, the dataset includes both SMILES and SELFIES representations. SMILES (Simplified Molecular Input Line Entry System) and SELFIES (SELF-referencIng Embedded Strings) are two popular notations used to represent chemical structures in a compact and standardized format. The dataset was created with the aim of training a predictive model, specifically using the Molecular Transformer architecture.

bioprocessing, hydrotreating, deep learning algori↗

DOE OSTI2020

Microgrid Energy Management System Integration with Advanced Distribution Management System

The Integrated Distribution Management System (IDMS) project was initiated to demonstrate the interactive operation of microgrid systems and the distribution systems with which they interconnect. The key technologies for this are the microgrid management system and the utility distribution management system. The IDMS project successfully demonstrated that a utility’s advanced distribution management system/distributed energy resources management system (ADMS/DERMS) could effectively manage microgrids to provide visibility and control functionalities as well as use the microgrid as a dispatchable resource to support the utility grid. The DERMS accomplishes this by determining active and reactive power needs at the point of common coupling (PCC) using advanced applications like volt/VAR watt optimization (VVWO) and load relief (LR) with the underlying core applications state estimation (SE) and load flow (LF). The ADMS/DERMS can use the microgrid as a resource to resolve and prevent violations in the grid and to optimize the operational working state of the grid. The IDMS project demonstrated that a utility-operated ADMS with embedded DERMS functionality can flexibly manage a variety of microgrids and other aggregated distributed energy resources (DER) in concert with the wider distribution grid. Microgrids can provide grid services in any number of different ways to meet the operational needs of the distribution utility. The manner of aggregation — microgrid or virtual power plant — is not necessarily relevant to the utility as long as the grid services from the aggregated DER are available and can be managed by its ADMS/DERMS for the stability and reliability of the grid. The IDMS project demonstrated this integrated ADMS concept by combining hardware/software-in-the-loop testing with commercial products from different vendors and a utility’s network model. The project integrated the Schneider Electric EcoStruxure™ ADMS with DERMS with the Schweitzer Engineering Laboratories (SEL) POWERMAX ® Microgrid Control System, and the simulated resources with energy company PECO’s model of a utility-owned microgrid, establishing an operational relationship in which the utility manages the operational functionality of a microgrid at the PCC and that provides the utility with the capability to control the comprehensive power system, inclusive of the macrogrid and microgrid.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2021

A multi-dimensional parametric study of variability in multi-phase flow dynamics during geologic CO 2 sequestration accelerated with machine learning

Successful geologic CO 2 storage projects depend on numerical simulations to predict reservoir performance during site selection, injection verification, and post-injection monitoring phases of the project. These numerical simulations solve non-linear sets of coupled partial differential equations, while accounting for multi-phase fluid dynamics on the basis of constitutive equations that are embedded into the solution scheme. As a consequence, individual simulations often require tens to hundreds of hours to complete on high-performance computing clusters. Moreover, laboratory experiments reveal that parametric functions for capillary pressure and relative permeability exhibit substantial variability, even within the same rock type. This combination of computational expense and wide-ranging parametric variability means that there remains substantial uncertainty in the behavior of multi-phase CO 2 -water systems, particularly in the context of feedbacks between relative permeability and capillary pressure. To bridge this knowledge gap, here we develop a novel workflow that utilizes physics-based numerical simulation to train an artificial neural network (ANN) emulator for interrogating the multivariate parameter space that governs both capillary pressure and relative permeability. With this approach, the ANN is trained to emulate both fluid pressure distribution and CO 2 saturation, which are then interrogated quantitatively to generate parametric response surface mappings with high-fidelity resolution. Results from this study initially show that capillary entry pressure is the dominant control on both CO 2 plume geometry and fluid pressure propagation when considering the combined effects of capillary pressure and relative permeability, particularly when phase interference is low and residual CO 2 saturation is high. Moreover, the ANN emulator provides tremendous computational speed-up by computing 2691 individual simulations in several minutes; whereas, the same simulation ensemble would have required ~3 years of simulation time using only physics-based simulation methods (25,000 times speed up).

58 GEOSCIENCES↗

DOE OSTI2025

Active deep kernel learning of molecular properties from structural embeddings

As vast databases of chemical identities become increasingly available, the challenge shifts to how we effectively explore and leverage these resources to study molecular properties. This paper presents an active learning approach for molecular discovery using deep kernel learning (DKL), demonstrated on the QM9 dataset. DKL links structural embeddings directly to properties, creating organized latent spaces that prioritize relevant property information. By iteratively recalculating embedding vectors in alignment with target properties, DKL uncovers concentrated maxima representing key molecular properties and reveals unexplored regions with potential for innovation. This approach underscores DKL’s potential in advancing molecular research and discovery.

Artificial neural networks↗

DOE OSTI2021

A generalized machine learning workflow to visualize mechanical discontinuity

Accurate detection and mapping of mechanical discontinuity in materials has widespread industrial and research applications. Herein, we developed a generalized machine-learning framework for visualizing single mechanical discontinuity embedded in material of any composition, velocity, density, porosity, and size with limited data. The proposed visualization of discontinuity requires accurate estimations of the length, location, and orientation of the embedded discontinuity by processing multipoint wave-transmission measurements. k-Wave simulator is used to create a large dataset of elastic waveforms recorded during multi-point wave-transmission measurements through materials containing single mechanical discontinuity. k-Wave simulator considers the wave attenuation, dispersion, and mode conversion in wave motion. Discrete wavelet transform (DWT) and statistical feature extraction are essential for data preprocessing prior to the data-driven model development. DWT also minimizes the effect of noise. Using hyper-parameter tuning and cross validation, gradient boosting regression can visualize the mechanical discontinuity with an accuracy of 0.85, in terms of coefficient of determination. A double-layered neural network-based regression has better performance with an accuracy of 0.95. Use of convolutional neural network converts the predictive task from a waveform processing to an image processing problem. Convolutional neural network achieved a generalization performance of 0.91. The proposed generalized workflow requires robust simulation of wave propagation, signal processing, feature engineering, and model evaluation. Sensors closest to the source and those located opposite the source are the most significant for the desired visualization. Notably, the sensors closest to the source capture the non-linear associations, whereas the sensor on the border opposite to the source capture the linear associations between the measured waveforms and the properties of the mechanical discontinuity.

42 ENGINEERING↗

DOE OSTI2023

Faster approximate subgraph counts with privacy

One of the most common problems studied in the context of differential privacy for graph data is counting the number of non-induced embeddings of a subgraph in a given graph. These counts have very high global sensitivity. Therefore, adding noise based on powerful alternative techniques, such as smooth sensitivity and higher-order local sensitivity have been shown to give significantly better accuracy. However, all these alternatives to global sensitivity become computationally very expensive, and to date efficient polynomial time algorithms are known only for few selected subgraphs, such as triangles, k-triangles, and k-stars. In this paper, we show that good approximations to these sensitivity metrics can be still used to get private algorithms. Using this approach, we much faster algorithms for privately counting the number of triangles in real-world social networks, which can be easily parallelized. We also give a private polynomial time algorithm for counting any constant size subgraph using less noise than the global sensitivity; we show this can be improved significantly for counting paths in special classes of graphs

Nguyen, Dung↗

DOE OSTI2022

Precoder Design for Physical-Layer Security and Authentication in Massive MIMO UAV Communications

Supporting reliable and seamless wireless connectivity for unmanned aerial vehicles (UAVs) has recently become a critical requirement to enable various different use cases of UAVs. Due to their widespread deployment footprint, cellular networks can support beyond visual line of sight (BVLOS) communications for UAVs. In this paper, we consider cellular connected UAVs (C-UAVs) that are served by massive multiple input-multiple-output (MIMO) links to extend coverage range, while also improving physical layer security and authentication. Here, we consider Rician channel and propose a novel linear precoder design for transmitting data and artificial noise (AN). We derive the closed-form expression of the ergodic secrecy rate of CUAVs for both conventional and proposed precoder designs. In addition, we obtain the optimal power splitting factor that divides the power between data and AN by asymptotic analysis. Then, we apply the proposed precoder design in the fingerprint embedding authentication framework, where the goal is to minimize the probability of detection of the authentication tag at an eavesdropper. In simulation results, we show the superiority of the proposed precoder in both secrecy rate and the authentication probability considering moderate and large number of antenna massive MIMO scenarios.

99 GENERAL AND MISCELLANEOUS↗

DOE OSTI2021

Gaseous cyclodextrin- closo -dodecaborate complexes χCD·B 12 X 12 2– (χ = α, β, and γ; X = F, Cl, Br, and I): electronic structures and intramolecular interactions

A fundamental understanding of cyclodextrin- closo -dodecaborate inclusion complexes is of great interest in supramolecular chemistry. Herein, we report a systematic investigation on the electronic structures and intramolecular interactions of perhalogenated closo -dodecaborate dianions B 12 X 12 2– (X = F, Cl, Br and I) binding to α-, β-, and γ-cyclodextrins (CDs) in the gas phase using combined negative ion photoelectron spectroscopy (NIPES) and density functional theory (DFT) calculations. The vertical detachment energy (VDE) of each complex and electronic stabilization of each dianion due to the CD binding (ΔVDE, relative to the corresponding isolated B 12 X 12 2– ) are determined from the experiments along α-, β- and γ-CD in the form of VDE (ΔVDE): 4.00 (2.10), 4.33 (2.43), and 4.30 (2.40) eV in X = F; 4.09 (1.14), 4.64 (1.69), and 4.69 (1.74) eV in X = Cl; 4.11 (0.91), 4.58 (1.38), and 4.70 (1.50) eV in X = Br; and 3.54 (0.74), 3.88 (1.08), and 4.05 (1.25) eV in X = I, respectively. All complexes have significantly higher VDEs than the corresponding isolated dodecaborate dianions with ΔVDE spanning from 0.74 eV at (α, I) to 2.43 eV at (β, F), sensitive to both host CD size and guest substituent X. DFT-optimized complex structures indicate that all B 12 X 12 2– prefer binding to the wide openings of CDs with the insertion depth and binding motif strongly dependent on the CD size and halogen X. Dodecaborate anions with heavy halogens, i.e. , X = Cl, Br, and I, are found outside of α-CD, while B 12 F 12 2– is completely wrapped by γ-CD. Partial embedment of B 12 X 12 2– into CDs is observed for the other complexes via multipronged B–X … H–O/C interlocking patterns. The simulated spectra based on the density of states agree well with those of the experiments and the calculated VDEs well reproduce the experimental trends. Molecular orbital analyses suggest that the spectral features at low binding energies originated from electrons detached from the dodecaborate dianion, while those at higher binding energies are derived from electron detachment from CDs. Energy decomposition analyses reveal that the electrostatic interaction plays a dominating role in contributing to the host–guest interactions for the X = F series partially due to the formation of a O/C–H … X–B hydrogen bonding network, and the dispersion forces gradually become important with the increase of halogen size.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Convergence of micro-geochemistry and micro-geomechanics towards understanding proppant shale rock interaction: A Caney shale case study in southern Oklahoma, USA

As a direct outcome of economic development coupled with an increase in population, global energy demand will continue to rise in the coming decades. Although renewable energy sources are increasingly investigated for optimal production, the immediate needs require focus on energy sources that are currently available and reliable, with a minimal environmental impact; the efficient exploration and production of unconventional hydrocarbon resources is bridging the energy needs and energy aspirations, during the current energy transition period. The main challenges are related to the accurate quantification of the critical rock properties that influence production, their heterogeneity and the multiscale driven physico-chemical nature of rock–fluid interactions. A key feature of shale reservoirs is their low permeability due to dominating nanoporosity of the clay-rich matrix. As a means of producing these reservoirs in a cost-effective manner, a prerequisite is creation of hydraulic fracture networks capable of the highest level of continued conductivity. Fracturing fluid chemical design, formation brine geochemical composition, and rock mineralogy all contribute to swelling-induced conductivity damage. The Caney Shale is an organic-rich, often calcareous mudrock. Many studies have examined the impact that clay has on different kinds of shale productivity but there is currently no data reported on the Caney Shale in relation to horizontal drilling; all reported data on the Caney Shale is on vertical wells which are shallow, compared to an emerging play that is at double the depth. Here, in this work we develop geochemical–geomechanical integration of rock properties at micro-and nanoscales that can provide insights into the potential proppant embedment and its mitigation. The novel methodology amalgamates the following: computed X-ray tomography, scanning electron microscopy, energy dispersive spectroscopy, micro-indentation, and Raman spectroscopy techniques. Here, our results show that due to the multiscale heterogeneity in the Caney Shale, these geochemical and structural properties translate into a variation in mechanical properties that will impact interaction between the proppant and the host shale rock.

03 NATURAL GAS↗

DOE OSTI2023

Coupled Long-Term Experiment and Model Investigation of the Differential Response of Plants and Soil Microbes in a Changing Permafrost Tundra Ecosystem

The major research goal of this project was to understand and quantify the fate of carbon stored in permafrost ecosystems using a combination of field and laboratory experiments to measure and model isotope ratios and carbon fluxes in a tundra ecosystem exposed to experimental warming. Field measurements centered on a two-factor experimental warming to increase air and soil temperatures alone, and in combination, at a tundra field site in Alaska. A second manipulation of water table was embedded within the warming treatment such that both major environmental factors (temperature, moisture) controlling ecosystem carbon dynamics were experimentally altered. The experiment was interfaced with modeling activities using both data assimilation and forward modeling approaches. Models were used to make forecasts of ecosystem carbon dynamics beyond the time frame and environmental space of the experiment itself. This project was a follow-on to previous work that made use of the same field manipulation; here we continued the experimental manipulation in order to test new hypotheses regarding long-term effects of warming and permafrost thaw and subsequent changes in moisture availability on ecosystem carbon dynamics. These results were linked with ongoing synthesis and model intercomparison activities through the Permafrost Carbon Network, which led to additional synthesis publications as a result of collaboration and data sharing from this project. As key outcomes, this project has produced new papers published in the peer-reviewed literature, archived datasets used for model-data intercomparisons, trained graduate students and postdoctoral researchers, and has raised awareness about the vulnerability of permafrost carbon to a wider audience.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

QC-GN 2 oMS 2 : a Graph Neural Net for High Resolution Mass Spectra Prediction

Predicting the mass spectrum of a molecular ion is often accomplished via three generalized approaches: rules-based methods for bond breaking, deep learning, or quantum chemical (QC) modeling. Rules-based approaches are often limited by the conditions for different chemical subspaces and perform poorly under chemical regimes with few defined rules. QC modeling is theoretically robust but requires significant amounts of computational time to produce a spectrum for a given target. Among deep learning techniques, graph neural networks (GNNs) have performed better than previous work with fingerprint-based neural networks in mass spectra prediction. To explore this technique further, we investigate the effects of including quantum chemically derived information as edge features in the GNN to increase predictive accuracy. The models we investigated include categorical bond order, bond force constants derived from extended tight-binding (xTB) quantum chemistry, and acyclic bond dissociation energies. Throughout this work, we evaluated these models against a control GNN with no edge features in the input graphs. Bond dissociation enthalpies yielded the best improvement with a cosine similarity score of 0.462 relative to the baseline model (0.437). In this work we also apply dynamic graph attention which improves performance on benchmark problems and supports the inclusion of edge features. Between implementations, we investigate the nature of the molecular embedding for spectra prediction and discuss the recognition of fragment topographies in distinct chemistries for further development in tandem mass spectrometry prediction.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Embedded Error Bayesian Calibration of Thermal Decomposition of Organic Materials

Organic materials are an attractive choice for structural components due to their light weight and versatility. However, because they decompose at low temperatures relative to tradiational materials they pose a safety risk due to fire and loss of structural integrity. To quantify this risk, analysts use chemical kinetics models to describe the material pyrolysis and oxidation using thermogravimetric analysis. This process requires the calibration of many model parameters to closely match experimental data. Previous efforts in this field have largely been limited to finding a single best-fit set of parameters even though the experimental data may be very noisy. Furthermore the chemical kinetics models are often simplified representations of the true de- composition process. The simplification induces model-form errors that the fitting process cannot capture. In this work we propose a methodology for calibrating decomposition models to thermogravimetric analysis data that accounts for uncertainty in the model-form and experimental data simultaneously. The methodology is applied to the decomposition of a carbon fiber epoxy composite with a three-stage reaction network and Arrhenius kinetics. The results show a good overlap between the model predictions and thermogravimetric analysis data. Uncertainty bounds capture devia- tions of the model from the data. The calibrated parameter distributions are also presented. In conclusion, the distributions may be used in forward propagation of uncertainty in models that leverage this material.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Self-Supervised T-GCN for Detection of Disturbance and Propagation in Power Grid

Urban power systems increasingly rely on dense sensing to monitor grid reliability, yet disturbance labels are scarce and events are rare. We present a self-supervised spatio-temporal method that detects, localizes, and characterizes grid frequency disturbances across urban areas using only unlabeled data. Our approach trains a tiny Temporal Graph Convolutional Network (T-GCN) to forecast per-site frequency residuals (deviation from 60 Hz). The sensor graph is constructed directly from signals using pre-event Pearson correlation with a cross-correlation lag penalty without geocoding. At inference, node-level anomalies are the model's forecast errors; region-level alarms arise from connected components of high-score nodes. We estimate disturbance propagation by computing per-node arrival times (first persistent exceedance), then fit a planar or time-of-arrival model to obtain direction, speed, and an epicenter proxy. With only three real events collected at decisecond resolution across U.S. cities, we evaluate the T-GCN and report time-to-detect, footprint size, and propagation consistency. We further show that short-window embeddings from the T-GCN's hidden states enable few-shot event-vs-background recognition via a simple prototypical classifier. Despite minimal data and no labels, our system yields fast, spatially coherent detection and interpretable propagation maps, offering a practical, lightweight pathway to city-scale grid resilience analytics.

Niu, Haoran [ORNL] (ORCID:0000000155228297)↗

DOE OSTI2025

Physics-informed latent neural operator for real-time predictions of time-dependent parametric PDEs

Deep operator network (DeepONet) has shown significant promise as surrogate models for systems governed by partial differential equations (PDEs), enabling accurate mappings between infinite-dimensional function spaces. However, when applied to systems with high-dimensional input-output mappings arising from large numbers of spatial and temporal collocation points, these models often require heavily overparameterized networks, leading to long training times. Latent DeepONet addresses some of these challenges by introducing a two-step approach: first learning a reduced latent space using a separate model, followed by operator learning within this latent space. While efficient, this method is inherently data-driven and lacks mechanisms for incorporating physical laws, limiting its robustness and generalizability in data-scarce settings. Here, in this work, we propose PI-Latent-NO, a physics-informed latent neural operator framework that integrates governing physics directly into the learning process. Our architecture features two coupled DeepONets trained end-to-end: a Latent-DeepONet that learns a low-dimensional representation of the solution, and a Reconstruction-DeepONet that maps this latent representation back to the physical space. By embedding PDE constraints into the training via automatic differentiation, our method eliminates the need for labeled training data and ensures physics-consistent predictions. The proposed framework is both memory and compute-efficient, exhibiting near-constant scaling with problem size and demonstrating significant speedups over traditional physics-informed operator models. We validate our approach on a range of parametric PDEs, showcasing its accuracy, scalability, and suitability for real-time prediction in complex physical systems.

Latent representations↗