Search NASA⌕ Search

SEARCH · Search NASA

Results for “reduced set methods”

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 199 records · Page 11
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Uncertainty Quantification Enabled by Automatic Differentiation for Hydrodynamic Simulation of Shock‐to‐Detonation Transition in High Explosives

Quantifying the effects of uncertainty in a reactive burn model on the run-to-detonation time in high explosives (HEs) provides a robust methodology for assessing the probability of an HE failing the IHE qualification standard. Moreover, uncertainty quantification helps evaluate whether the model calibration accurately represents data outside the calibration set. This study uses a specialized hydrodynamic simulation code for modeling detonation to determine the run-to-detonation time of the HE PBX 9502 for various impact velocities. To quickly approximate uncertainties in the model, a surrogate was constructed using a Taylor series expansion centered at the mean of the input parameters. To obtain the sensitivities required for constructing the Taylor series, HYP-percomplex Automatic Differentiation (HYPAD) was implemented. HYPAD is a methodology for infusing existing codes with automatic differentiation capabilities by augmenting variables with one or more imaginary units to compute step-size independent partial derivatives. These derivatives are accurate to machine precision with respect to the implemented numerical algorithm, meaning their accuracy reflects that of the underlying method (e.g., integration or discretization schemes). Using reduced order modeling techniques, the mean and standard deviation of the run-to-detonation time of a shock within PBX 9502 were computed for a number of initial impact velocities. A weighted least squares regression was then performed to obtain a best fit curve and prediction interval for the computed statistics. Historical data points from explosively driven wedge tests were utilized to validate the prediction interval, ensuring its reliability in predicting future outcomes. With this prediction interval and a known safety constraint curve, the most probable point of failure and the probability of failure for the HE PBX 9502 were determined.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Solving high-dimensional partial integral differential equations: The finite expression method

Partial integro-differential equations (PIDEs) have broad applications in the sciences, from electro-magnetism to options pricing. Here, in this paper, we introduce a new finite expression method (FEX) to solve PIDEs. This approach builds upon the original FEX and its inherent advantages with new advances: 1) A novel method of parameter grouping is proposed to reduce the number of coefficients in high-dimensional function approximation; 2) A Taylor series approximation method is implemented to significantly improve the computational efficiency and accuracy of the evaluation of the integral terms of PIDEs. The new FEX based method, denoted FEX-PG to indicate the addition of the parameter grouping (PG) step to the algorithm, provides both high accuracy and interpretable numerical solutions, with the outcome being an explicit equation that facilitates intuitive understanding of the underlying solution structures. These features are often absent in traditional methods, such as finite element methods (FEM) and finite difference methods, as well as in deep learning-based approaches. To benchmark our method against recent advances, we apply the new FEX-PG to solve benchmark PIDEs in the literature. In high-dimensional settings, FEX-PG exhibits strong and robust performance, achieving relative errors on the order of single precision machine epsilon, significantly outperforming existing approaches based on neural networks.

Combinatorial optimization↗
DOE OSTI2025

Toward Drilling the Perfect Geothermal Well: An International Research Coordination Network for Geothermal Drilling Optimization Supported by Deep Machine Learning and Cloud Based Data Aggregation

The EDGE project, supported by the U.S. Department of Energy Geothermal Technologies Office under award DE-EE0008793, established a data-driven framework for improving the efficiency, cost-effectiveness, and reliability of geothermal well drilling. The project focused on developing scalable data infrastructure, advanced machine learning and probabilistic models, and integrated analytics tools to support continuous drilling optimization. A central objective was to reduce geothermal drilling costs by up to seventy percent while minimizing the risk of well failure through predictive diagnostics and adaptive planning. Over the project period, a comprehensive data repository was designed and deployed, incorporating records from over one hundred geothermal wells across varied geological settings. This repository supported both structured and unstructured data and adhered to FAIR data principles, enabling provenance tracking, quality control, and standardized metadata. The project introduced automated ingestion pipelines and a cloud-hosted platform that facilitated access to raw, processed, and derived datasets. This infrastructure served as the foundation for model development and analysis. Machine learning workflows were developed to predict key drilling metrics including rate of penetration, non-productive time, and total drilling costs. Self-organizing maps and dimensionality reduction methods were used to uncover operational patterns and outliers, while supervised learning algorithms such as random forests and deep neural networks were applied to forecast performance outcomes. The models were validated on heterogeneous datasets from both U.S. and Icelandic fields, demonstrating variable but significant predictive accuracy. The results indicated that finer temporal resolution, inclusion of lithological data, and consistency in operational annotations could substantially improve model performance. The project also implemented process mining techniques to reconstruct state-transition models from drilling event logs. These models enabled the identification of deviations from optimal workflows and provided insights into recurring failure modes. Analysis of non-productive time highlighted the impact of equipment failures, geological challenges, and human factors, offering opportunities for targeted mitigation strategies. The EDGE Dashboard was developed as a web-based expert system integrating data visualization, model outputs, and user-driven queries. It provided an accessible interface for operators to explore historical data, evaluate predicted outcomes, and compare drilling scenarios. Initial feedback from project partners suggested that the dashboard could serve as a foundation for more advanced advisory and optimization tools. Overall, the EDGE project demonstrated the feasibility and value of applying modern data science techniques to geothermal drilling. It delivered a set of interoperable tools and models that can support more efficient, lower-risk well development. The findings point toward a viable path for transitioning from advisory analytics to semi-autonomous drilling systems, contingent on continued collaboration, expanded datasets, and field validation. The project results have immediate relevance for drilling operations, data management practices, and future geothermal R&D efforts aimed at achieving reliable, cost-competitive geothermal energy at scale.

15 GEOTHERMAL ENERGY↗
DOE OSTI2025

Physics-driven Explosions of Stripped High-mass Stars: Synthetic Light Curves and Spectra of Stripped-envelope Supernovae with Broad Light Curves

Stripped-envelope supernovae (SESNe) represent a significant fraction of core-collapse supernovae, arising from massive stars that have shed their hydrogen and, in some cases, helium envelopes. The origins and explosion mechanisms of SESNe remain a topic of active investigation. In this work, we employ radiative-transfer simulations to model the light curves and spectra of a set of explosions of single, solar-metallicity, massive Wolf–Rayet stars with ejecta masses ranging from 4 to 11 M ⊙ , which were computed from a turbulence-aided and neutrino-driven explosion mechanism. We analyze these synthetic observables to explore the impact of varying ejecta mass and helium content on observable features. We find that the light curve shape of these progenitors with high ejecta masses is consistent with observed SESNe with broad light curves but not the peak luminosities. The commonly used analytic formula based on rising bolometric light curves overestimates the ejecta mass of these high-initial-mass progenitor explosions by a factor of up to 2.6. In contrast, the calibrated method by Haynie et al., which relies on late-time decay tails, reduces uncertainties to an average of 20% within the calibrated ejecta mass range. Spectroscopically, the He I 1.083 μm line remains prominent even in models with as little as 0.02 M ⊙ of helium. However, the strength of the optical He I lines is not directly proportional to the helium mass but instead depends on a complex interplay of factors such as the 56 Ni distribution, composition, and radiation field. Thus, producing realistic helium features requires detailed radiative transfer simulations for each new hydrodynamic model.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

Antarctic ice sheet model comparison with uncurated geological constraints shows that higher spatial resolution improves deglacial reconstructions

Accurately reconstructing past changes to the shape and volume of the Antarctic ice sheet relies on the use of physically based and thus internally consistent ice sheet modeling, benchmarked against spatially limited geologic data. The challenge in model benchmarking against geologic data is diagnosing whether model-data misfits are the result of an inadequate model, inherently noisy or biased geologic data, and/or incorrect association between modeled quantities and geologic observations. In this work we address this challenge by (i) the development and use of a new model-data evaluation framework applied to an uncurated data set of geologic constraints, and (ii) nested high-spatial-resolution modeling designed to test the hypothesis that model resolution is an important limitation in matching geologic data. While previous approaches to model benchmarking employed highly curated datasets, our approach applies an automated screening and quality control algorithm to an uncurated public dataset of geochronological observations (specifically, cosmogenic-nuclide exposure-age measurements from glacial deposits in ice-free areas). This optimizes data utilization by including more geological constraints, reduces potential interpretive bias, and allows unsupervised assimilation of new data as they are collected. We also incorporate a nested model framework in which high-resolution domains are downscaled from a continent-wide ice sheet model. We highlight the application of this framework by applying these methods to a small ensemble of deglacial ice-sheet model simulations, and demonstrate that the nested approach improves the ability of model simulations to match exposure age data collected from areas of complex topography and ice flow. We develop a range of diagnostic model-data comparison metrics to provide more insight into model performance than possible from a single-valued misfit statistic, showing that different metrics capture different aspects of ice sheet deflation.

Geosciences↗
DOE OSTI2024

Inefficacy of mallard flight responses to approaching vehicles

Vehicle collisions with birds are financially costly and dangerous to humans and animals. To reduce collisions, it is necessary to understand how birds respond to approaching vehicles. We used simulated (i.e., animals exposed to video playback) and real vehicle approaches with mallards (Anas platyrynchos) to quantify flight behavior and probability of collision under different vehicle speeds and times of day (day vs. night). Birds exposed to simulated nighttime approaches exhibited reduced probability of attempting escape, but when escape was attempted, fled with more time before collision compared to birds exposed to simulated daytime approaches. The lower probability of flight may indicate that the visual stimulus of vehicle approaches at night (i.e., looming headlights) is perceived as less threatening than when the full vehicle is more visible during the day; alternatively, the mallard visual system might be incompatible with vehicle lighting in dark settings. Mallards approached by a real vehicle exhibited a delayed margin of safety (both flight initiation distance and time before collision decreased with speed); they are the first bird species found to exhibit this response to vehicle approach. Our findings suggest mallards are poorly equipped to adequately respond to fast-moving vehicles and demonstrate the need for continued research into methods promoting effective avian avoidance behaviors.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Measurements of soil protist richness and community composition are influenced by primer pair, annealing temperature, and bioinformatics choices

ABSTRACT Protists are a diverse and understudied group of microbial eukaryotic organisms especially in terrestrial environments. Advances in molecular methods are increasing our understanding of the distribution and functions of these creatures; however, there is a vast array of choices researchers make including barcoding genes, primer pairs, PCR settings, and bioinformatic options that can impact the outcome of protist community surveys. Here, we tested four commonly used primer pairs targeting the V4 and V9 regions of the 18S rRNA gene using different PCR annealing temperatures and processed the sequences with different bioinformatic parameters in 10 diverse soils to evaluate how primer pair, amplification parameters, and bioinformatic choices influence the composition and richness of protist and non-protist taxa using Illumina sequencing. Our results showed that annealing temperature influenced sequencing depth and protist taxon richness for most primer pairs, and that merging forward and reverse sequencing reads for the V4 primer pairs dramatically reduced the number of sequences and taxon richness of protists. The data sets of primers that targeted the same 18S rRNA gene region (e.g., V4 or V9) had similar protist community compositions; however, data sets from primers targeting the V4 18S rRNA gene region detected a greater number of protist taxa compared to those prepared with primers targeting the V9 18S rRNA region. There was limited overlap of protist taxa between data sets targeting the two different gene regions (80/549 taxa). Together, we show that laboratory and bioinformatic choices can substantially affect the results and conclusions about protist diversity and community composition using metabarcoding. IMPORTANCE Ecosystem functioning is driven by the activity and interactions of the microbial community, in both aquatic and terrestrial environments. Protists are a group of highly diverse, mostly unicellular microbes whose identity and roles in terrestrial ecosystem ecology have been largely ignored until recently. This study highlights the importance of choices researchers make, such as primer pair, on the results and conclusions about protist diversity and community composition in soils. In order to better understand the roles protist taxa play in terrestrial ecosystems, biases in methodological and analytical choices should be understood and acknowledged.

Biotechnology & Applied Microbiology↗
DOE OSTI2026

Performing Numerical Analysis of Cybersecurity Options Using Dynamic Risk Analysis Tool EMRALD

Cyberattacks can have many different attack paths, durations, and goals. There are also many different mitigation options involving hardware, software, and/or humans. Considering a cyber threat should involve defense-in-depth methods and a quantitative or numerical evaluation of overall effectiveness against dynamic, time-dependent attacks to make cost and risk-informed decisions. Typical cyberattack modeling methods only provide a qualitative evaluation. The main areas of cybersecurity are confidentiality, integrity, and availability. For companies with cyber-physical systems such as advanced nuclear reactors, cyber-related safety is a requirement set by North American Electric Reliability and the U.S. Nuclear Regulatory Commission. They are also concerned about availability or reliability as a business case. As cyber threats are evolving to a business-for-hire structure, more attacks may focus on disrupting business success and reliability, causing financial and economic stability risk. Companies want to know business reliability and recovery from those threats, and that requires modeling physical behavior of the targets. Dynamic-state-based and Markov-based modeling provides a method for better cyber scenario modeling with different tools having issues such as state-base explosion. Dynamic modeling enables time and conditional features not found in other numerical evaluation methods. EMRALD (Event Modeling Risk Assessment using Lined Diagrams) is a dynamic risk analysis modeling and simulation tool and has features that reduce modeling issues. It has been used to model different time-dependent events including plant behavior and operator procedures. As a general modeling tool, EMRALD can also be used to model cyberattack scenarios with varying mitigation options and quantify effectiveness, producing numerical data for risk-informed decisions. This paper uses EMRALD to demonstrate that dynamic numerical risk analysis can be used for cyber threat modeling to provide insights for design decision-making and optimize defense strategies. Keywords: cyber modeling; cyber-physical systems; numerical cyber modeling

97 - MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Field Validation of MVA Technology for Offshore CCS: Novel Ultra-High-Resolution 3D Marine Seismic Technology (P-Cable) (Final Report)

The objectives of the proposed study were to deploy and validate a specific monitoring technology, high-resolution 3D marine seismic (HR3D), appropriate for large-demonstration and commercial-scale offshore CCS sites. The project accomplished successful acquisition two HR3D seismic surveys. The first HR3D dataset was over the offshore injection site of the Tomakomai, Japan integrated pilot CCS project, which at the time of survey acquisition was actively injecting CO 2 . The first survey also represented a successful international collaboration between the DOE NETL program and Japan’s national CCS program and was the first successful acquisition and use of HR3D over an active CO 2 injection site (Meckel, Feng et al. 2019). The Tomakomai HR3D survey successfully tested a novel 4-streamer HR3D system array in which, for the first time, no cross-cable (aka “P-Cable”) was utilized and only four GeoEel streamers were used instead of the standard 12-streamer configuration. Consequently, this was not, strictly speaking, a deployment of the “P-Cable” system of (Planke and Berndt 2004) but rather a modified version, thereof, and it is the first known demonstration of the modified system configuration. One very positive outcome from the Japanese collaboration earlier in the project was the ability to learn from the Japanese how they used tail buoys with GPS to determine the position of the seismic source and receivers in time and space. Based on that experience, GCCC designed and built six GPS receivers that could be used to position the streamer receivers and the seismic source via tail buoys. A fundamental advance that was made on the original design, was the ability to directly power the tail buoy GPS units and transfer data through the streamers (i.e., vs. the batteries used at Tomakomai). The bulkiness of the GPS batteries caused drag and episodic surging of the buoys, which affected data quality by lifting up the tail end of the streamers so the receivers were not at the same depth. The units were tested onshore for accuracy and functionality, and the design was subsequently and successfully tested in marine acquisition mode during the SLP survey acquisition. The marine acquisition test and survey satisfied Subtasks 2.2.2, Novel Positioning Technology Selection and Subtask 2.2.3, Novel Positioning Technology Deployment. Results of the novel positioning technology selection (Subtask 2.2.2) were considered successful and will be incorporated in future HR3D seismic acquisition projects to reduce costs, improve deployment safety at sea, and integrate both seismic and data recording via a single data transfer through the streamers to the recording system. The project also established a permitting process through NETL NEPA compliance, which included an Environmental Assessment in a marine setting and is required for conducting these types of surveys using Federal funding. The permitting process charted a “boilerplate,” which can allow future surveys related to other funded projects to move forward more expeditiously. Future improvements that could be considered are more robust seals on the GPS module and stronger materials (especially joints) on tail buoy fabrication. These would increase fixed costs, but would be advisable and probably more economic long-term if multiple HR3D surveys are planned. Project Accomplishments include: • Pre-survey Sensitivity Study • Marine geochemistry methods and data analysis • Successful HR3D seismic dataset acquired @ Tomakomai active CO 2 injection marine site • Developed advanced seismic processing techniques • No NRMS anomalies detected in overburden; Demonstration of containment • Repeatability study • Second survey collected @ San Luis Pass, TX • 4D application using positioning techniques developed in the project for monitoring were successful

3D seismic GPS positioning↗
DOE OSTI2026

Product Defect Detection System: SYSM- 5620 Final Project

Retail sales is a growing market estimated to up to seven percent year over year. With this growing market there is also a trend in growing rate of retail returns, estimated just last year at $\$$850 billion. Retail stores must ensure that products available for purchase remain safe, undamaged, and acceptable to customers throughout their time in the store. This job exists regardless of the specific solution used because stores are always responsible for preventing damaged or defective products from reaching customers and when they fail to this is categorized under operation inefficiencies which accounts for an estimated $\$$12 billion in returns. When defective items remain on the sales floor, stores may experience increased returns, reduced customer satisfaction, loss of customer trust, and potential safety concerns depending on the product type. As a result, the core job to be done is to identify defective products quickly, remove them from the sales floor before they are purchased, and preserve useful information about the defect so that the store can improve its handling, stocking, and supplier coordination over time. The need for a more reliable process is especially important in high volume retail environments where employees manage large numbers of products across many aisles, shelves, and storage areas. In these settings, manual inspection alone can be inconsistent and difficult to sustain at the individual item level. At the same time, broader retail trends continue to emphasize operational efficiency, product visibility, and improved customer experience, creating an opportunity for more automated and data driven defect detection methods.

42 ENGINEERING↗
DOE OSTI2026

Thermodynamic Control of Interface Directs MnO 2 Nucleation Chemistry for Dense and Conformal Electrodeposition

Manganese dioxide (MnO 2 ) is widely recognized as a promising material for high-energy-density energy storage systems due to its broad applicability and facile electrodeposition. However, achieving uniform, thin, and high-mass-loading MnO 2 coatings on high-surface-area electrodes remains a significant challenge. Conventional electrodeposition methods typically yield nonuniform, thick layers with poor conductivity and limited material utilization, restricting their practical use. Here, we uncover a thermodynamically engineered vanadyl/pervanadyl (VO 2+ /VO 2 + ) interface that fundamentally reshapes MnO 2 electrodeposition chemistry, enabling highly uniform and dense coatings. Here, combining in situ AFM measurement, Classical Nucleation Theory, and Johnson–Mehl–Avrami–Kolmogorov modeling, we show that this interface reduces early-stage detectable MnO 2 island size by 35-fold and shifts the MnO 2 growth from diffusion-limited to reaction-limited progressive nucleation. This thermodynamically controlled interface yields highly dense and conformal MnO 2 films with record-high mass loading of 241 mg cm –2 (1607 mg cm –3 ) on 3D-printed graphene aerogels, without compromising porosity or inducing thickness gradient. As a prototype demonstration, the resulting MnO 2 electrodes deliver record-setting volumetric performance in both capacitors (106 F cm –3 ) and Zn//MnO 2 pouch cells (162 mAh cm –3 ). Beyond energy storage, our findings demonstrate the significance of thermodynamic interface control in MnO 2 nucleation chemistry for achieving dense and uniform coatings on various substrates, with implications for electrocatalysis, semiconductor processing, and advanced materials manufacturing.

Batteries↗
DOE OSTI2024

Multitask methods for predicting molecular properties from heterogeneous data

Data generation remains a bottleneck in training surrogate models to predict molecular properties. We demonstrate that multitask Gaussian process regression overcomes this limitation by leveraging both expensive and cheap data sources. In particular, we consider training sets constructed from coupled-cluster (CC) and density functional theory (DFT) data. We report that multitask surrogates can predict at CC-level accuracy with a reduction in data generation cost by over an order of magnitude. Of note, our approach allows the training set to include DFT data generated by a heterogeneous mix of exchange–correlation functionals without imposing any artificial hierarchy on functional accuracy. More generally, the multitask framework can accommodate a wider range of training set structures—including the full disparity between the different levels of fidelity—than existing kernel approaches based on Δ-learning although we show that the accuracy of the two approaches can be similar. Consequently, multitask regression can be a tool for reducing data generation costs even further by opportunistically exploiting existing data sources.

Chemistry↗
DOE OSTI2026

Analyzing users’ preferences between personal and pooled rideshare services using a mixed logit modeling approach

Ridesharing has become an increasingly popular transportation method over the past decade. Transportation network companies such as Uber and Lyft generally provide two types of rideshare services: personal rideshare, in which users ride alone or with individuals they know, and pooled rideshare, in which users ride with passengers they do not know but share similar routes. Pooled rideshare is capable of reducing energy consumption and traffic in the transportation system in comparison to personal rideshare. Despite the growth in trip volume, ridesharing usage is still low compared to other popular transportation methods in the U.S., particularly traveling in one’s own personal vehicle. Furthermore, pooled rideshare usage is lower than personal rideshare. To understand riders’ preferences, a national survey (N = 2884) was conducted in the U.S. to investigate users’ choice behaviors in rideshare services examining personal versus pooled rideshare. Each survey respondent completed 20 stated-preference scenarios where participants choose between a personal or pooled rideshare option. Based on the responses, a mixed logit model was developed to capture the choice behavior preferences of the participants. The model unveiled the impact of demographic and trip attribute variables on users’ rideshare preferences. The discussion encompassed insights into demographic backgrounds and trip attributes, accompanied by a set of policy recommendations aimed at enhancing future pooled rideshare utilization.

Choice model↗
DOE OSTI2025

Toward a Balanced Description of Ground and Excited States with Transcorrelated F12 Methods

By correlating only the 1-particle states occupied in the reference determinant, the conventional design for the single-reference R12/F12 explicitly correlated methods biases them toward the ground-state description, thereby making the treatment of response properties of the ground state, and energies and other properties of excited states less robust. While the use of multireference methods and/or extensions of the standard SP-projected geminals can achieve a more balanced description of ground and excited states, here we show that the same goals can be achieved by extending the action of F12 correlators to the occupied and valence unoccupied 1-particle states only. This design choice reflects the strong dependence of the optimal correlation length scale of the F12 ansatz on the orbital energies/structure, and helps to avoid the unphysical raising of the ground-state energy if the F12 geminals are used to correlate pairs of all 1-particle states. The improved F12 geminal design is incorporated into the unitary transcorrelation framework to produce a unitary 2-body Hamiltonian that incorporates the short-range dynamical correlation physics for ground and low-energy excited states in a balanced manner. This explicitly correlated effective Hamiltonian reduces the basis set requirement on the correlation-consistent basis cardinal number by 1 or more over the uncorrelated counterpart for the description of ground-state coupled-cluster singles and doubles (CCSD) energies, vertical excitation energies, and harmonic vibrational frequencies of equation-of-motion CCSD low-energy excited states.

Hamiltonians↗
DOE OSTI2025

Attention-based functional-group coarse-graining: a deep learning framework for molecular prediction and design

Machine learning (ML) offers considerable promise for the design of new molecules and materials. In real-world applications, the design problem is often domain-specific, and suffers from insufficient data, particularly labeled data, for ML training. In this study, we report a data-efficient, deep-learning framework for molecular discovery that integrates a coarse-grained functional-group representation with a self-attention mechanism to capture intricate chemical interactions. Our approach exploits group-contribution concepts to create a graph-based intermediate representation of molecules, serving as a low-dimensional embedding that substantially reduces the data demands typically required for training. Using a self-attention mechanism to learn the subtle but highly relevant chemical context of functional groups, the method proposed here consistently outperforms existing approaches for predictions of multiple thermophysical properties. In a case study focused on adhesive polymer monomers, we train on a limited dataset comprising only 6,000 unlabeled and 600 labeled monomers. The resulting chemistry prediction model achieves over 92% accuracy in forecasting properties directly from SMILES strings, exceeding the performance of current state-of-the-art techniques. Furthermore, the latent molecular embedding is invertible, enabling the design pipeline to automatically generate new monomers from the learned chemical subspace. We illustrate this functionality by targeting several properties, including high and low glass transition temperatures (Tg), and demonstrate that our model can identify new candidates with values that surpass those in the training set. The ease with which the proposed framework navigates both chemical diversity and data scarcity offers a promising route to accelerate and broaden the search for functional materials.

Han, Ming [Univ. of Chicago, IL (United States)]↗
DOE OSTI2026

Differentiable multiphase flow model for physics-informed machine learning in reservoir pressure management

Accurate subsurface reservoir pressure control is extremely challenging due to geological heterogeneity and multiphase fluid-flow dynamics. Predicting behavior in this setting relies on high-fidelity physics-based simulations that are computationally expensive. Yet, the uncertain, heterogeneous properties that control these flows make it necessary to perform many of these expensive simulations, which is often prohibitive. To address these challenges, we introduce a physics-informed machine learning workflow that couples a fully differentiable multiphase flow simulator, which is implemented in the DPFEHM framework with a convolutional neural network (CNN). The CNN learns to predict fluid extraction rates from heterogeneous permeability fields to enforce pressure limits at critical reservoir locations. By incorporating transient multiphase flow physics into the training process, our method enables more practical and accurate predictions for realistic injection-extraction scenarios compared to previous works. To speed up training, we pretrain the model on single-phase, steady-state simulations and then finetune it on full multiphase scenarios, which dramatically reduces the computational cost. We demonstrate that high-accuracy training can be achieved with fewer than three thousand full-physics multiphase flow simulations – compared to previous estimates requiring up to ten million. This drastic reduction in the number of simulations is achieved by leveraging transfer learning from much less expensive single phase simulations.

25 ENERGY STORAGE↗
DOE OSTI2025

Simulating Atmospheric Processes in Earth System Models and Quantifying Uncertainties With Deep Learning Multi‐Member and Stochastic Parameterizations

Abstract Deep learning is a powerful tool to represent subgrid processes in climate models, but many application cases have so far used idealized settings and deterministic approaches. Here, we develop stochastic parameterizations with calibrated uncertainty quantification to learn subgrid convective and turbulent processes and surface radiative fluxes of a superparameterization embedded in an Earth System Model (ESM). We explore three methods to construct stochastic parameterizations: (a) a single Deep Neural Network (DNN) with Monte Carlo Dropout; (b) a multi‐member parameterization; and (c) a Variational Encoder Decoder with latent space perturbation. We show that the multi‐member parameterization improves the representation of convective processes, especially in the planetary boundary layer, compared to individual DNNs. The respective uncertainty quantification illustrates that methods (b) and (c) are advantageous compared to a dropout‐based DNN parameterization regarding the spread of convective processes. Hybrid simulations with our best‐performing multi‐member parameterizations remained challenging and crash within the first days. Therefore, we develop a pragmatic partial coupling strategy relying on the superparameterization for condensate emulation. Partial coupling reduces the computational efficiency of hybrid Earth‐like simulations but enables model stability over 5 months with our multi‐member parameterizations. However, our hybrid simulations exhibit biases in thermodynamic fields and differences in precipitation patterns. Despite this, the multi‐member parameterizations enable improvements in reproducing tropical extreme precipitation compared to a traditional convection parameterization. Despite these challenges, our results indicate the potential of a new generation of multi‐member machine learning parameterizations leveraging uncertainty quantification to improve the representation of stochasticity of subgrid effects.

Behrens, Gunnar [Deutsches Zentrum für Luft‐ und R↗
DOE OSTI2025

Investigation of the Effect of Gate Oxide Screening with Adjustment Pulse on Commercial SiC Power MOSFETs

This paper presents a method to recover the negative threshold voltage shift during high field gate oxide screening of 1.2 kV 4H-SiC MOSFETs with an additional adjustment gate voltage pulse. To reduce field failure rates of the MOSFETs in operation, manufacturers perform a screening treatment to remove devices with extrinsic defects in the oxide. Current gate oxide screening procedures are limited to oxide fields at or below ~9 MV/cm for short durations (<1 s), which is not enough to remove all the devices with extrinsic defects. The results show that by implementing a lower field gate pulse, the threshold voltage shift can be partially recovered, and therefore the maximum screening field and time can be increased. However, both the initial screening pulse and the adjustment pulse require careful calibration to prevent significant degradation of the device threshold voltage, on-resistance, interface state density, or intrinsic lifetime. With a well calibrated set of pulses, higher screening fields can be utilized without significantly damaging the devices. This leads to an improvement in the overall screening efficiency of the process, reducing the number of devices with extrinsic oxide defects entering the field, and improving the reliability of the SiC MOSFETs in operation.

42 ENGINEERING↗