Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “prediction”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Subseasonal Forecasting and MJO Teleconnections in Machine Learning Weather Prediction Models

Abstract In recent years, machine‐learning (ML) models trained on reanalysis data have rivaled physics‐based forecast models in terms of performance skill for global weather forecasting. With increased rollout stability, the question of how these models perform for subseasonal to seasonal (S2S, week 3–8) forecasting has emerged. In this study we run a large set of subseasonal hindcasts over 2004–2023 to evaluate two ML weather forecast models at the S2S time scale, SFNO‐HENS (Nvidia, fully ML) and NeuralGCM (Google Research, hybrid). Corresponding hindcasts from the European Centre for Medium‐Range Weather Forecasts (ECMWF) are used as a baseline for comparison to a physics‐based model. Because our focus is on predicting moisture transport over the Western United States between October and March, we evaluate the models' prediction skill for the Madden‐Julian Oscillation (MJO) and its associated teleconnections in the North Pacific. We find that both ML models are competitive with the ECWMF model, with comparable skill in predicting the North Pacific large‐scale circulation and the MJO at week 3 and beyond. Even though overall the mid‐latitude subseasonal prediction skill remains low, the ML models exhibit interesting behavior such as a realistic propagation of the MJO across the Maritime Continent and realistic teleconnections. A SFNO‐HENS sensitivity experiment with altered initial conditions in the tropics demonstrates the stability of the model, and it illustrates the capability of ML models to represent important physical processes of the atmosphere at the S2S time scale. Plain Language Summary Predicting weather patterns and precipitation a few weeks in advance (subseasonal time scale) is of great interest for stakeholders such as water managers in the Southwest United States (US), where arid conditions prevail. Subseasonal forecasts from traditional weather forecast models exhibit low skill in the region, limiting their applicability. Here we examine whether the recent breakthrough in weather forecasting made with machine learning/artificial intelligence models can translate to improved subseasonal forecasts. Recently‐developed machine learning models exhibit comparable skill to a state‐of‐the‐art physics‐based model for predicting weather patterns in the North Pacific/North America region, and associated moisture transport. The same applies to their skill in predicting the tropical pattern, the Madden‐Julian Oscillation, and its important remote perturbations over the midlatitude East Pacific and Southwest US. Additionally, a perturbation experiment carried out with one of the machine learning models illustrates their ability to not only predict the evolution of atmospheric fields, but also to learn and represent physical processes such as tropics‐extratropics Rossby wave propagation. Key Points Two machine learning weather forecast models exhibit state‐of‐the‐art prediction skill at the subseasonal time scale in the Pacific sector The models equal ECWMF in terms of Madden‐Julian oscillation (MJO) prediction skill, and they accurately predict the MJO propagation and associated teleconnections The two machine‐learning models represent key physical processes for subseasonal prediction, despite being trained for weather forecasting

Peings, Yannick↗
DOE OSTI2026

Benchmarking core turbulence and transport predictions for an inductive compact tokamak reactor plasma

Motivated by the need for accurate, timely, and efficient calculations of plasma transport, predictions of plasma turbulence properties made using different TGLF saturation rules are benchmarked against corresponding predictions from linear and nonlinear gyrokinetic CGYRO simulations. This benchmarking is carried out using parameters taken from an inductive burning plasma scenario in a hypothetical compact high-field (R maj = 4 m, B T = 8 T) tokamak, lying in a much different regime of parameter space than either the TGLF calibration regime or current-day experiments. The core turbulent transport in this scenario is predicted to be dominated by ion temperature gradient (ITG) turbulence. In general, the ITG critical gradients predicted by various TGLF saturation rules are quite close to the CGYRO predictions. Both codes predict similar linear ITG growth rates and frequency spectra, as well as their scaling with R/L T i = −Rd ln(T i )/dr. However, TGLF systematically predicts unstable trapped-electron modes (TEMs) above k y ρ s ≃ 0.5 not seen by CGYRO for the same parameters, due to TGLF predicting a lower threshold in R/L T e than CGYRO for TEM onset. It is shown that for this scenario, nonlinear CGYRO simulations predict stiffer ITG turbulence than the TGLF SAT0 and SAT1 saturation rules, with energy fluxes close in magnitude and scaling with R/L T i to what is predicted by the SAT2 saturation rule. Self-consistent core profiles calculated using nonlinear CGYRO flux predictions and the PORTALS transport solver are shown to agree fairly well with corresponding predictions made using the TGLF SAT2 model, including a similar level of density peaking.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Tandem Predictions for HPC Jobs: Preprint

At the core of the predictive analytics applied to High Performance Computing (HPC), the most prominent tasks are the prediction of job runtimes and the prediction of job queue times, both of which have the potential for informing HPC users during their every-day decision making. Accurate runtime predictions can help users better choose so-called wallclock times at job submission, decreasing the odds of their jobs waiting in queues longer than necessary. The accurate and timely queue time predictions offered for the available partitions can inform the favorable selection of partitions for running jobs. This potential is well understood as we see in the abundance of research studies that propose solutions for these tasks, including the work published in the last several years. These tasks are seemingly receptive to the Machine Learning (ML) solutions, considering that there is no shortage of training data where HPC centers over time run millions and millions of jobs. However, we study the existing research literature, as well as look for examples in the toolchains supported on the exemplar HPC facilities, and, surprisingly, do not find any practical solutions that are ready to be adopted. We interpret this as a manifestation of the shortage of UX/UI efforts that support HPC analytics and also as a sign that the research has not come to the consensus on solving these tasks. In this study, we aim to shed new light on the long-running task of job queue time prediction by exploring the utility of runtime predictions in improving prediction accuracy and, actually, predicting these two metrics together, in tandem. In other words, we show how runtime predictions become valuable input in the queue time modeling. We challenge the existing approaches to feature engineering for the queue time prediction and describe promising results we obtained for a large dataset of HPC jobs from a supercomputer at the National Renewable Energy Laboratory.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Unveiling the Potential of MeshGraphNets for Predicting Subsurface Evolution in Carbon Storage Projects

This is the conference paper accompanying an oral presentation “Unveiling the Potential of MeshGraphNets for Predicting Subsurface Evolution in Carbon Storage Projects” at the 17th International Conference on Greenhouse Gas Control Technologies GHGT-17 held in Calgary, Canada, October 20-24 , 2024. Carbon capture and storage (CCS) technology is critical for mitigating climate change but requires effective subsurface reservoir management to ensure safe containment of injected CO2. Accurate predictions of reservoir pressure and saturation are essential for assessing long-term CCS performance. Traditional numerical simulations, while effective, are computationally intensive, time-consuming, and constrained by data discretization. Previous work has shown the effectiveness of MeshGraphNets (MGN), a graph-based machine learning framework, as an innovative alternative for predicting reservoir behavior. MGN leverages graph neural networks (GNNs) and mesh representations to model complex geological formations, offering superior adaptability across different discretizations and reservoir configurations. Classic MGN implementations utilize an autoregressive technique to predict future behavior based on current predictions, but this technique is hampered by error accumulation over time. To enhance the model accuracy in time-series predictions, this study implemented a multi-step rollout strategy that integrates autoregressive predictions during training to stabilize prediction of saturation over time. Using the Illinois Basin – Decatur Project (IBDP) dataset, comprising 100 simulations of CO2 injection, pressure, and saturation changes, the framework demonstrated its ability to learn spatial dependencies and temporal dynamics. With inputs including permeabilities, porosities, and injection rates, MGN accurately predicted CO2 plume evolution over time, even with limited training data. Moreover, the addition of a multi-step rollout procedure during training improved the ability of MGN to predict stably over time by ~15%. This research positions MGN, enhanced with multi-step rollout capabilities, as a robust and efficient tool for CCS applications. It advances the field by enabling precise, computationally efficient predictions of reservoir behavior, providing a foundation for the broader adoption of machine learning frameworks in CCS and other geoscience domains.

Holcomb, Paul↗
DOE OSTI2024

Comparison of Machine Learning-Based Predictive Models of the Nutrient Loads Delivered from the Mississippi/Atchafalaya River Basin to the Gulf of Mexico

Predicting nutrient loads is essential to understanding and managing one of the environmental issues faced by the northern Gulf of Mexico hypoxic zone, which poses a severe threat to the Gulf’s healthy ecosystem and economy. The development of hypoxia in the Gulf of Mexico is strongly associated with the eutrophication process initiated by excessive nutrient loads. Due to the complexities in the excessive nutrient loads to the Gulf of Mexico, it is challenging to understand and predict the underlying temporal variation of nutrient loads. The study was aimed at identifying an optimal predictive machine learning model to capture and predict nonlinear behavior of the nutrient loads delivered from the Mississippi/Atchafalaya River Basin (MARB) to the Gulf of Mexico. For this purpose, monthly nutrient loads (N and P) in tons were collected from US Geological Survey (USGS) monitoring station 07373420 from 1980 to 2020. Machine learning models—including autoregressive integrated moving average (ARIMA), gaussian process regression (GPR), single-layer multilayer perceptron (MLP), and a long short-term memory (LSTM) with the single hidden layer—were developed to predict the monthly nutrient loads, and model performances were evaluated by standard assessment metrics—Root Mean Square Error (RMSE) and Correlation Coefficient (R). The residuals of predictive models were examined by the Durbin–Watson statistic. The results showed that MLP and LSTM persistently achieved better accuracy in predicting monthly TN and TP loads compared to GPR and ARIMA. In addition, GPR models achieved slightly better test RMSE score than ARIMA models while their correlation coefficients are much lower than ARIMA models. Moreover, MLP performed slightly better than LSTM in predicting monthly TP loads while LSTM slightly outperformed for TN loads. Furthermore, it was found that the optimizer and number of inputs didn’t show effects on the LSTM performance while they exhibited impacts on MLP outcomes. This study explores the capability of machine learning models to accurately predict nonlinearly fluctuating nutrient loads delivered to the Gulf of Mexico. Further efforts focus on improving the accuracy of forecasting using hybrid models which combine several machine learning models with superior predictive performance for nutrient fluxes throughout the MARB.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Machine Learning Framework for Conotoxin Class and Molecular Target Prediction

Conotoxins are small and highly potent neurotoxic peptides derived from the venom of marine cone snails which have captured the interest of the scientific community due to their pharmacological potential. These toxins display significant sequence and structure diversity, which results in a wide range of specificities for several different ion channels and receptors. Despite the recognized importance of these compounds, our ability to determine their binding targets and toxicities remains a significant challenge. Predicting the target receptors of conotoxins, based solely on their amino acid sequence, remains a challenge due to the intricate relationships between structure, function, target specificity, and the significant conformational heterogeneity observed in conotoxins with the same primary sequence. We have previously demonstrated that the inclusion of post-translational modifications, collisional cross sections values, and other structural features, when added to the standard primary sequence features, improves the prediction accuracy of conotoxins against non-toxic and other toxic peptides across varied datasets and several different commonly used machine learning classifiers. Here, we present the effects of these features on conotoxin class and molecular target predictions, in particular, predicting conotoxins that bind to nicotinic acetylcholine receptors (nAChRs). We also demonstrate the use of the Synthetic Minority Oversampling Technique (SMOTE)-Tomek in balancing the datasets while simultaneously making the different classes more distinct by reducing the number of ambiguous samples which nearly overlap between the classes. In predicting the alpha, mu, and omega conotoxin classes, the SMOTE-Tomek PCA PLR model, using the combination of the SS and P feature sets establishes the best performance with an overall accuracy (OA) of 95.95%, with an average accuracy (AA) of 93.04%, and an f1 score of 0.959. Using this model, we obtained sensitivities of 98.98%, 89.66%, and 90.48% when predicting alpha, mu, and omega conotoxin classes, respectively. Similarly, in predicting conotoxins that bind to nAChRs, the SMOTE-Tomek PCA SVM model, which used the collisional cross sections (CCSs) and the P feature sets, demonstrated the highest performance with 91.3% OA, 91.32% AA, and an f1 score of 0.9131. The sensitivity when predicting conotoxins that bind to nAChRs is 91.46% with a 91.18% sensitivity when predicting conotoxins that do not bind to nAChRs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Multi-trait multi-environment genomic prediction strategies for Miscanthus sacchariflorus

Genomic selection holds the potential to serve as a strategic tool to enhance the genetic gain of complex traits in Miscanthus breeding programs. The development of improved cultivars requires their assessment for various traits across diverse environments to ensure suitable overall performance. Hence, the multi-trait multi-environment (MTME) genomic prediction (GP) models offer an opportunity to improve selection accuracy. This study aims to evaluate the potential of five GP models: (1) three MTME models including genotype-by-trait-by-environment interaction (G×E×T) and (2) two single-trait multi-environment (STME) models (with and without G×E interaction). A Miscanthus sacchariflorus population comprising 336 genotypes evaluated in three environments and scored for four traits (biomass yield YDY, total culm number TCM, average internode length AIL, and culm node number CNN) was analyzed. The predictive ability of the models was evaluated considering three cross-validation schemes resembling realistic scenarios (CV1: predicting new genotypes, CVP: predicting missing traits in a given environment, and CV2: predicting partially observed genotypes). On average, in all cross-validation schemes compared to the STME the predictive ability of the MTME models was 10% to 70% higher for TCM and AIL. On the other hand, for YDY and CNN, both STME models performed similarly or slightly better (between 5 to 64%) than the MTME models in most environments. While the MTME models were not successful for all traits when compared to their STME counterparts, MTME models improved the prediction of the performance of genotypes that were untested across environments or lacked trait information in a specific environment. Overall, our study suggests that MTME GP models can be implemented in Miscanthus breeding programs to improve the predictive ability of the complex traits, shorten breeding cycles, and accelerate selection decisions.

genomic prediction (GP)↗
DOE OSTI2026

Evaluation of methods and improvement of predictions for specification properties of petroleum-based and alternative aviation fuels

To support our research and process modeling for liquid fuels, including blends, from petroleum and synthetic sources such as from biomass intermediates, we evaluated composition-based prediction methods and improved predictions for five key specification properties of petroleum-based and alternative aviation fuels, namely distillation temperatures (10 % distilled, t 10 , and final boiling point, t FBP ), density, flash point, net heat of combustion, and freezing point. The types of fuels included were petroleum-based jet fuels, jet-fuel surrogate mixtures, synthetic blending components obtained from different sources, and blends of Jet A with many synthetic blending components. Expanded datasets to update associated parameters allowed significant improvements for one of the prediction methods used in earlier work, namely the Modified Weighted Average method published initially by Shi et al. By considering the importance of lighter compounds for flash points and heavier compounds for freezing points, the revised Modified Weighted Average method was further improved. For liquid density, the revised Modified Weighted Average method gave the best overall results. The revised Modified Weighted Average method, the American Society for Testing and Materials D7215 method, and the D7215 method modified by another group gave comparable results for flash point, while the revised Modified Weighted Average and D3338 methods gave the best results for net heat of combustion. Freezing point was well predicted using the revised Modified Weighted Average method and showed the most significant improvements over current predictions. Distillation temperature t 10 was not well predicted, while t FBP was predicted with a mean absolute error comparable to experimental reproducibility.

09 BIOMASS FUELS↗
DOE OSTI2024

Data-driven prediction of scaling and ignition of inertial confinement fusion experiments

Recent advances in inertial confinement fusion (ICF) at the National Ignition Facility (NIF), including ignition and energy gain, are enabled by a close coupling between experiments and high-fidelity simulations. Neither simulations nor experiments can fully constrain the behavior of ICF implosions on their own, meaning pre- and postshot simulation studies must incorporate experimental data to be reliable. Linking past data with simulations to make predictions for upcoming designs and quantifying the uncertainty in those predictions has been an ongoing challenge in ICF research. We have developed a data-driven approach to prediction and uncertainty quantification that combines large ensembles of simulations with Bayesian inference and deep learning. The approach builds a predictive model for the statistical distribution of key performance parameters, which is jointly informed by past experiments and physics simulations. The prediction distribution captures the impact of experimental uncertainty, expert priors, design changes, and shot-to-shot variations. We have used this new capability to predict a 10× increase in ignition probability between Hybrid-E shots driven with 2.05 MJ compared to 1.9 MJ, and validated our predictions against subsequent experiments. We describe our new Bayesian postshot and prediction capabilities, discuss their application to NIF ignition and validate the results, and finally investigate the impact of data sparsity on our prediction results.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

GRUMDN: A Multi-Task Model for Predicting Human Patterns-of-Life from Stay Transition Data

Understanding human patterns-of-life (PoL) is essential towards ensuring safe and secure indoor facility environment as well as outdoor urban environment. Prediction of human movement in between places of interest is vital in understanding human PoL. Movement between spaces maybe represented and detected in one of the two forms: 1) trajectories: locations measured at regular time intervals by mobile sensors, bluetooth or GPS sensors; or 2) stay transitions: semantic PoI (points of interest) and stay duration data measurable by eventbased sensors that collect data when a check-in or check-out event is detected. Stay transition data provides a more compressed data format compared to trajectories data, especially in situations with longer stay durations, while preserving the information necessary for PoL analysis. Now as introduced briefly in the paper, our deployed end application (Digital Twin of a facility with non-player characters, besides the interactive user in virtual reality) needed a well-performing and validated AI/ML model for simulating high quality stay transitions behavior. In this study we thus primarily present our findings with developing and validating that model, which is a multi-task neural network for stay transition prediction. The neural network consists of two heads, for corresponding two tasks of stay category prediction and stay duration prediction. We evaluated gated recurrent units and multi-layer perceptrons of varying network sizes for stay category prediction; while mixture density networks, noisy generator-only networks, and generative adversarial networks of varying network sizes for stay duration prediction. We have then evaluated four multi-task models, constructed by combining these specialized models, on their ability to predict stay transition data. We tested our models on datasets from two different cases: 1) a simulation-generated dataset of indoor movement within the HFIR (high flux isotope reactor) nuclear reactor facility at Oak Ridge National Laboratory (ORNL); and 2) the GeoLife human mobility dataset of outdoor urban movement available in literature. Our results indicate that GRUMDN, which combines gated recurrent units (GRU) for stay category prediction task, and mixture density networks (MDN) for stay duration prediction task, did overall outperform other multitask models and the current state-of-the-art.

Gunaratne, Chathika [ORNL] (ORCID:0000000225088745↗
DOE OSTI2025

A Digital Twin Framework Utilizing Machine Learning for Robust Predictive Maintenance: Enhancing Tire Health Monitoring

We introduce a novel digital twin (DT) framework for the predictive maintenance of long-term physical systems. Using monitoring tire health as an application, we show how the DT framework can be used to enhance automotive safety and efficiency, and how the technical challenges can be overcome using a three-step approach. First, to manage the data complexity over a long operation span, we employ data reduction techniques to concisely represent physical tires using historical performance and usage data. Relying on these data, for fast real-time prediction, we train a transformer-based model offline on our concise dataset to predict future tire health over time, represented as remaining casing potential (RCP). Based on our architecture, our model quantifies both epistemic and aleatoric uncertainties, providing reliable confidence intervals around predicted RCP. Second, to incorporate real-time data, we update the predictive model in the DT framework, ensuring its accuracy throughout its lifespan with the aid of hybrid modeling and the use of the discrepancy function. Third, to assist decision-making in predictive maintenance, we implement a tire state decision algorithm, which strategically determines the optimal timing for tire replacement based on RCP forecasted by our transformer model. This approach ensures that our DT accurately predicts system health, continually refines its digital representation, and supports predictive maintenance decisions. Furthermore, our framework effectively embodies a physical system, leveraging big data and machine learning (ML) for predictive maintenance, model updates, and decision-making.

advanced computing infrastructure↗
DOE OSTI2025

Evaluation of New Additions to OLI Software in Predicting Mercuric and Mercurous Species in Liquid Waste Operations

Speciation of mercury during the pretreatment steps of tank waste processing is critical to successful mercury removal prior to vitrification during Liquid Waste Operations (LWO) at SRS. OLI software has been used to predict mercury speciation and activity throughout LWO. The OLI software operates based on a thermodynamic framework called the Mixed Solvent Electrolyte (MSE) framework. The MSE framework allows prediction in theoretically infinitely dilute to concentrated mixtures (e.g., purely solute solutions). Before modification to the MSE framework databanks, certain critical mercury species were missing in the MSE databank, and some thermodynamic data needed to be updated for the OLI software to accurately predict mercury chemical species in SRS waste tanks. To better reflect streams across LWO, new mercury species were integrated into the MSE database. To evaluate the changes to the OLI MSE framework per the Technical Task Request (TTR) and the Task Technical and Quality Assurance Plan (TTQAP), waste stream compositions from Tanks 38, 43, and Tank 50 decontaminated salt solution (DSS) were used as model inputs. Models were developed and executed using both the old and new databases. Compositional analyses from caustic Tank 50 DSS and caustic Tanks 38 and 43 were used as the input streams. These streams represent the most comprehensive chemical data sets where both mercury and tank constituents were measured together. Results for Tank 50 DSS predict HgO as the predominant species in both databases. Both methyl and dimethyl Hg species are present when the new database is ‘on’ and are not predicted with the new database turned ‘off’. The new database predicts a greater amount of HgO and a greater fraction of it in the solid phase. Pourbaix diagrams (potential vs. pH) generated for each Tank 50 DSS were identical regardless of which database was used. Elemental Hg and HgO were predicted in the water stable region under basic conditions. Tanks 38 and 43 follow similar trends as the Tank 50 DSS models. Unlike Tanks 38 and 50 DSS, the Tank 43 Pourbaix plot shows a region of stability for an aqueous HgOHCO3 - species between approximately pH 7-11. In all streams, when MeHg+ is included in the inputs, the new database predicts aqueous MeHgOH as the dominant species. If elemental or dimethyl mercury is in the waste stream, the new database model predicts they are unchanged and remain in those states and quantities. Additionally, the total mercury values are reported for both the measured input data and the OLI output data for all considered tanks. The summary indicates that the percentage error between the measured and calculated values is less than 1% in all cases The reconciliations and generation of the Pourbaix diagrams for Tank 50 DSS took approximately ten times longer with the new database ‘on’. In addition, over the course of that time, models with the new database ‘on’ were more likely to crash or display an error. Some modest performance improvements were noted when modeling with an i7 processor versus an i5. An example error is found in Appendix A. Furthermore, Appendix B provides V&V for two chemical systems analyzed with the OLI software, results were satisfactory. It is recommended to utilize the new databases (i.e., HCO.ddb and SR-Hg.ddb) in future Savannah River Mission Completion applications of OLI to represent pseudo steady-state. Furthermore, the integration and utilization of the new databases (i.e., HCO.ddb and SR-Hg.ddb) in modeling applications (e.g., Aspen) is also recommended.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Predictability of the early summer surface air temperature over Western South Asia

Variability of the Surface Air Temperature (SAT) over the Western South Asia (WSA) region leads to frequent heatwaves during the early summer (May-June) season. The present study uses the European Centre for Medium-Range Weather Forecast’s fifth-generation seasonal prediction system, SEAS5, from 1981 to 2022 based on April initial conditions (1-month lead) to assess the SAT predictability during early summer season. The goal is to evaluate the SEAS5’s ability to predict the El Niño-Southern Oscillation (ENSO) related interannual variability and predictability of the SAT over WSA, which is mediated through upper-level (200-hPa) geopotential height anomalies. This teleconnection leads to anomalously warm surface conditions over the region during the negative ENSO phase, as observed in the reanalysis and SEAS5. We evaluate SEAS5 prediction skill against two observations and three reanalyses datasets. The SEAS5 SAT prediction skill is higher with high spatial resolution observations and reanalysis datasets compared to the ones with low-resolution. Overall, SEAS5 shows reasonable skill in predicting SAT and its variability over the WSA region. Moreover, the predictability of SAT during La Niña is comparable to El Niño years over the WSA region.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

An investigation on machine learning predictive accuracy improvement and uncertainty reduction using VAE-based data augmentation

The confluence of ultrafast computers with large memory, rapid progress in Machine Learning (ML) algorithms, and the availability of large datasets place multiple engineering fields at the threshold of dramatic progress. However, a unique challenge in nuclear engineering is data scarcity because experimentation on nuclear systems is usually more expensive and time-consuming than most other disciplines. One potential way to resolve the data scarcity issue is deep generative learning, which uses certain ML models to learn the underlying distribution of existing data and generate synthetic samples that resemble the real data. In this way, one can significantly expand the dataset to train more accurate predictive ML models. In this study, our objective is to evaluate the effectiveness of data augmentation using variational autoencoder (VAE)-based deep generative models. We investigated whether the data augmentation leads to improved accuracy in the predictions of a deep neural network (DNN) model trained using the augmented data. Additionally, the DNN prediction uncertainties are quantified using Bayesian Neural Networks (BNN) and conformal prediction (CP) to assess the impact on predictive uncertainty reduction. To test the proposed methodology, we used TRACE simulations of steady-state void fraction data based on the NUPEC Boiling Water Reactor Full-size Fine-mesh Bundle Test (BFBT) benchmark. Here, we found that augmenting the training dataset using VAEs has improved the DNN model’s predictive accuracy, improved the prediction confidence intervals, and reduced the prediction uncertainties.

Bayesian neural network↗
DOE OSTI2025

Comparative Analysis of TCR and TCR-pMHC Complex Structure Prediction Tools

The rapid development of computational approaches for predicting the structures of T cell receptors (TCRs) and TCR-peptide-major histocompatibility (TCR-pMHC) complexes, accelerated by AI breakthroughs such as AlphaFold, has made it feasible to calculate these structures with increasing accuracy. Although these tools show great potential, their relative accuracy and limitations remain unclear due to the lack of standardized benchmarks. Here, we systematically evaluate seven tools for predicting isolated TCR structures together with six tools for predicting TCR-pMHC complex structures. The methods include homology-based approaches, general prediction tools using AlphaFold, TCR-specific tools derived from AlphaFold2, and the newly developed tFold-TCR model. The evaluation uses a post-training data set comprising 40 αβ TCRs and 27 TCR-pMHC complexes (21 Class I and 6 Class II). Model accuracy is assessed at global, local, and interface levels using a variety of metrics. We find that each tool offers distinct advantages in various aspects of its predictions. AlphaFold2, AlphaFold3, and tFold-TCR excel in overall accuracy of TCR structure prediction, and TCRmodel2 and AlphaFold2 perform well in overall accuracy of TCR-pMHC structure prediction. However, TCR-specific tools derived from AlphaFold2 show lower accuracy in the framework region than both homology-based methods and general-purpose tools such as AlphaFold, and challenges remain for all in modeling CDR3 loops, docking orientations, TCR-peptide interfaces, and Class II MHC-peptide interfaces. Furthermore, these findings will guide researchers in selecting appropriate tools, emphasize the importance of using multiple evaluation metrics to assess model performance, and offer suggestions for improving TCR and TCR-pMHC structure prediction tools.

Chemical structure↗
DOE OSTI2025

PySIDT: Subgraph Isomorphic Decision Trees for Molecular Property Prediction

Accurate molecular property prediction is important across all fields of chemistry. Deep neural networks (DNNs) have become increasingly popular due to their ability to train automatically, avoiding the incredibly tedious process of constructing and extending traditional property estimation schemes. However, DNNs require large amounts of training data, are challenging to interpret, require large amounts of memory to load even during inference, and have severe difficulties incorporating qualitative chemical knowledge, which are often desired for molecular property prediction tasks. Here, in this study, we present PySIDT (https://github.com/zadorlab/PySIDT), a software for training and running inference on Subgraph Isomorphic Decision Trees (SIDTs). SIDTs are graph-based decision trees made of nodes associated with molecular substructures. Inference is done by descending target molecular structures down the decision tree to nodes with matching subgraph isomorphic substructures and making predictions based on the final (most specific) nodes matched. SIDTs scale down well to dataset sizes much smaller than is feasible for DNNs. As trees of molecular substructures, SIDTs are inherently readable and easy to visualize, making them easy to analyze. They are also straightforward to extend and retrain, facilitate uncertainty estimation, and enable easy integration of expert knowledge. We demonstrate the SIDT approach discussing its application to a diverse range of molecular prediction tasks: rate coefficient estimation, diffusion coefficient estimation, thermochemistry estimation, transition state bond stretch prediction, p K a prediction, stability of molecular structures, stability of surface structures, and prediction of surface lateral interaction energetics. Additionally, we demonstrate the power of the SIDT algorithms in two direct learning curve vanilla comparisons with the popular DNN-based software Chemprop and the popular gradient boosted trees-based software XGBoost on enthalpy of formation and rate coefficient prediction tasks. In particular, in the enthalpy of formation case, vanilla PySIDT is able to outperform vanilla Chemprop and XGBoost across the full range of training/validation set sizes out to 11,560 data points.

Johnson, Matthew Sean [Sandia National Laboratorie↗
DOE OSTI2025

Sensitivities of time-dependent temperature profile predictions for NSTX with the multi-mode model

The Multi-Mode Model (MMM) for turbulent transport was applied to a large set of well-analyzed discharges from the National Spherical Torus Experiment (NSTX) in order to evaluate its sensitivities to a wide range of plasma conditions. MMM calculations were performed for hundreds of milliseconds in each discharge by performing time-dependent predictive simulations with the 1.5D tokamak integrated modeling code TRANSP. A closely related study (Lestz et al 2025 Plasma Phys. Control. Fusion 67 105029) concluded that MMM predicted electron and ion temperature profiles that were in reasonable agreement with NSTX observations, generally outperforming a different reduced transport model, TGLF. This finding motivates the more thorough investigation of the characteristics of the MMM predictions conducted in this work. The simulations with MMM have electron energy transport dominated by electron temperature gradient modes in the examined discharges with relatively low plasma β (ratio of kinetic plasma pressure to magnetic field pressure) and high collisionality, transitioning to a mixture of different modes for higher β and lower collisionality. The thermal ion diffusivity predicted by MMM is much smaller than the neoclassical contribution, in line with previous experimental analysis of NSTX. Nonetheless, the electron and ion temperature profiles are coupled via collisional energy exchange and thus sensitive to which transport channels are predicted. The time-dependent simulations with MMM are robust to the simulation start time, converging to remarkably similar temperature profiles later during the discharge. MMM typically overpredicts confinement relative to NSTX observations, leading to the prediction of overly steep temperature profiles. Plasmas with spatially broader temperature profiles, higher plasma β, and longer energy confinement times tend to be predicted by MMM with better agreement with the experiment. As a result, these findings provide useful context for understanding the regime-dependent tendencies of MMM in anticipation of self-consistent, time-dependent predictive simulations of NSTX-U discharges with these same modeling tools.

MMM↗
DOE OSTI2024

Evidential Deep Learning: Enhancing Predictive Uncertainty Estimation for Earth System Science Applications

Abstract Robust quantification of predictive uncertainty is a critical addition needed for machine learning applied to weather and climate problems to improve the understanding of what is driving prediction sensitivity. Ensembles of machine learning models provide predictive uncertainty estimates in a conceptually simple way but require multiple models for training and prediction, increasing computational cost and latency. Parametric deep learning can estimate uncertainty with one model by predicting the parameters of a probability distribution but does not account for epistemic uncertainty. Evidential deep learning, a technique that extends parametric deep learning to higher-order distributions, can account for both aleatoric and epistemic uncertainties with one model. This study compares the uncertainty derived from evidential neural networks to that obtained from ensembles. Through applications of the classification of winter precipitation type and regression of surface-layer fluxes, we show evidential deep learning models attaining predictive accuracy rivaling standard methods while robustly quantifying both sources of uncertainty. We evaluate the uncertainty in terms of how well the predictions are calibrated and how well the uncertainty correlates with prediction error. Analyses of uncertainty in the context of the inputs reveal sensitivities to underlying meteorological processes, facilitating interpretation of the models. The conceptual simplicity, interpretability, and computational efficiency of evidential neural networks make them highly extensible, offering a promising approach for reliable and practical uncertainty quantification in Earth system science modeling. To encourage broader adoption of evidential deep learning, we have developed a new Python package, Machine Integration and Learning for Earth Systems (MILES) group Generalized Uncertainty for Earth System Science (GUESS) (MILES-GUESS) ( https://github.com/ai2es/miles-guess ), that enables users to train and evaluate both evidential and ensemble deep learning. Significance Statement This study demonstrates a new technique, evidential deep learning, for robust and computationally efficient uncertainty quantification in modeling the Earth system. The method integrates probabilistic principles into deep neural networks, enabling the estimation of both aleatoric uncertainty from noisy data and epistemic uncertainty from model limitations using a single model. Our analyses reveal how decomposing these uncertainties provides valuable insights into reliability, accuracy, and model shortcomings. We show that the approach can rival standard methods in classification and regression tasks within atmospheric science while offering practical advantages such as computational efficiency. With further advances, evidential networks have the potential to enhance risk assessment and decision-making across meteorology by improving uncertainty quantification, a longstanding challenge. This work establishes a strong foundation and motivation for the broader adoption of evidential learning, where properly quantifying uncertainties is critical yet lacking.

Schreck, John S.↗