Results for “Task analysis” — Page 14

DOE OSTI2025

LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields

Model ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of per-sample errors obtained during training and employing a distance-based similarity search in the model latent space. Our method, which we call LTAU (Loss Trajectory Analysis for Uncertainty), efficiently estimates the full probability distribution function of errors for any test point using the logged training errors, achieving speeds that are 2–3 orders of magnitudes faster than typical ensemble methods and allowing it to be used for tasks where training or evaluating multiple models would be infeasible. We apply LTAU towards estimating parametric uncertainty in atomistic force fields (LTAU-FF), demonstrating that it produces well-calibrated confidence intervals and predicts errors that correlate strongly with the true errors for data near the training domain. Furthermore, we show that the errors predicted by LTAU-FF can be used in practical applications for detecting out-of-domain data, tuning model performance, and predicting failure during simulations. We believe that LTAU will be a valuable tool for uncertainty quantification in atomistic force fields and is a promising method that should be further explored in other domains of machine learning.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

AmpSuite

Seismic amplitudes offer vital information about explosion source characteristics, including discrimination and yield estimation. To take advantage of this, we developed an interactive Python package to measure, control data quality, generate broad area propagation models and perform discrimination and estimate yield. Propagation models are essential in support of transportable yield and broad area discrimination. The key benefit of this package will be its ability to continuously integrate data and new techniques. The AmpSuite framework will provide standardized, repeatable, and accurate model generation and characterization routines. The capability is crucial for monitoring agencies tasked with rapid and high-quality seismic event characterization. The AmpSuite software includes a series of independent modules to perform: • Direct Phase Amplitude Measurement and Storage • Coda Envelope Measurement and Storage • Data Quality Control • New Propagation Model Developments • Seismic Discrimination and Analysis • Yield Estimation and supporting utility software. The AmpSuite software provides comprehensive solutions for monitoring agencies seeking to optimize model generation and event analysis within a contemporary Python framework. Stakeholders (AFTAC) have begun to move towards the Python language for scientific analysis as a new workforce emerges.

Alfaro, Richard↗

DOE OSTI2025

VA EDH Advanced Software Pipeline Framework Report: Enhancing Automation and Scalability

The VA Environmental Determinants of Health (EDH) Advanced Software Pipeline Framework is designed to enhance the efficiency, scalability, and security of geospatial data processing workflows. This framework integrates modern data orchestration and containerization technologies, including Prefect for workflow automation, Docker for containerization, and PostgreSQL/PostGIS for geospatial data storage and analysis. It ensures standardized, reproducible, and automated data processing, supporting VA objectives related to substance use risk assessment and recovery research. The pipeline addresses key scalability and performance challenges through horizontal and vertical scaling, high-performance computing (HPC) integration, parallel processing, task caching, and dynamic resource allocation. These optimizations improve throughput and reduce latency, allowing the system to efficiently manage large and complex datasets. Additionally, security and compliance measures—such as data encryption (SSL), Role-Based Access Control (RBAC), and adherence to GDPR and HIPAA standards—safeguard sensitive information throughout data transmission and storage. A key implementation of this framework includes the automation of shelter list geolocation workflows, ensuring that up-to-date data is readily available for VA decision-making. Lessons learned from this project include the transition from in-memory processing to incremental storage writes, improving resource management and reliability. Future enhancements aim to expand automation, integrate AI-driven anomaly detection, and incorporate high-performance computing resources. This framework provides a scalable, secure, and adaptable solution for managing geospatial datasets, reinforcing the VA’s ability to support clinical and strategic initiatives through data-driven decision-making.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Nanoscopic Imaging of Self-Propelled Ultrasmall Catalytic Nanomotors

Ultrasmall nanomotors (<100 nm) are highly desirable nanomachines for their size-specific advantages over their larger counterparts in applications spanning nanomedicine, directed assembly, active sensing, and environmental remediation. While there are extensive studies on motors larger than 100 nm, the design and understanding of ultrasmall nanomotors have been scant due to the lack of high-resolution imaging of their propelled motions with orientation and shape details resolved. Here, we report the imaging of the propelled motions of catalytically powered ultrasmall nanomotors─hundreds of them─at the nanometer resolution using liquid-phase transmission electron microscopy. These nanomotors are Pt nanoparticles of asymmetric shapes (“tadpoles” and “boomerangs”), which are colloidally synthesized and observed to be fueled by the catalyzed decomposition of NaBH4 in solution. Statistical analysis of the orientation and position trajectories of fueled and unfueled motors, coupled with finite element simulation, reveals that the shape asymmetry alone is sufficient to induce local chemical concentration gradient and self-diffusiophoresis to act against random Brownian motion. Our work elucidates the colloidal design and fundamental forces involved in the motions of ultrasmall nanomotors, which hold promise as active nanomachines to perform tasks in confined environments such as drug delivery and chemical sensing.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2025

Comparison of Expert Vocabulary Usage Patterns Between Mental Health and Nonmental Health Clinicians When Diagnosing Pediatric Anxiety Disorders

Objective: To compare the utilization patterns of expert vocabulary (EVo) in diagnosing pediatric anxiety between mental health and non-mental health clinical notes from electronic health records to understand the role of Evo in informing classification and decision-making in anxiety diagnoses. Study design: We conducted a retrospective study using a cohort less than age 25 from Cincinnati Children's Hospital including 897 685 patients with 61 586 446 notes. We analyzed EVo, collected from mental health clinicians, in both mental and nonmental health notes. We compared classification accuracy using EVo-based patient-level embedding from all clinical notes, mental-health notes, and nonmental health notes for 2 tasks: 1) pre-vs postdiagnosis anxiety patients, and 2) prediagnosis anxiety vs nonanxiety patients. Results: EVo usage was highest in prediagnosis anxiety, lower in nonanxiety, and lowest in post-diagnosis. Classification models using EVo features from all, mental-health, and non-mental health notes showed similar F1 scores for prediagnosis anxiety (0.70 ± 0.2 for 2 categories). For anxiety vs nonanxiety classification, all clinical and nonmental health notes had better F1 scores than mental-health notes (above 0.90 for 3 categories). There was a notable difference in class-wise performance across both tasks. Conclusions: There are significant differences in anxiety EVo use between mental health and nonmental health clinicians. Despite less anxiety-specific terminology, non-mental health notes still captured key aspects of patient presentations, emphasizing the importance of including all clinicians' notes in analysis. EVo's utility for anxiety classification is most effective in prediagnostic phases, suggesting the need for a dedicated diagnostic lexicon and further study before incorporating EVo into classification models.

feature engineering↗

DOE OSTI2024

Northwest Combined Heat and Power Technical Assistance Partnership: Final Scientific/Technical Report

During the years of 2018 through 2023, The Washington State University Energy Program (WSU EP) operated the Northwest Combined Heat and Power Technical Assistance Partnership under contract to the US Department of Energy’s Advanced Manufacturing Office (AMO), a part of DOE’s larger Office of Energy Efficiency and Renewable Energy. This contract provided direction and funding for providing technical outreach, information and technical/economic analysis services in support of expanding development of Combined Heat and Power (CHP) over a four state region in the Pacific Northwest. The states served included Alaska, Idaho, Oregon and Washington. During the duration of this contract, the WSU EP delivered performed specific assigned tasks and provided services of various generally prescribed types, with the overall goal of increasing adoption of CHP as an efficient heat and power supply in the region. Indications – though not formally evaluated to our knowledge – are that this goal was achieved. Over many years, the WSU team has provided support for CHP developments in these states under similar previous Department of Energy contracts, and many such projects have been built. While the early-phase outreach, education and techno-economic assessments that the NW CHP TAP provides rarely immediately result in project construction, there is no doubt that the WSU EP team has had substantial influence in the construction of a number of important CHP projects in the region. Examples of projects implemented directly under the current team’s guidance and services during this contract cycle include an 875 kilowatt biomass fueled CHP system installed at the University of Idaho, and 5 megawatt natural gas fueled CHP system at the University of Montana ( UM Breaks Ground on New Power Facility, Slashes Emissions 30% (umt.edu) ). Many other CHP systems recently installed are also either directly or at least partially the result of information and technical services provided by the team. Finally, though much more difficult to quantify and less glamorous, the team has helped countless organizations to understand CHP and determine whether and how CHP may be a fit for them, before they consider paying for engineering services. Whether the ultimate outcome has been to include CHP or not, the expert skills provided by the NW CHP TAP team have provided real value to these organizations considering this relatively complex technology. Over the duration of the contract our team provided 71 Initial Technical Assessments and 17 Advanced Technical Assessments; 53 End-User Engagements and 33 Stakeholder Engagements of various types; 42 Technical Profiles (Project Profiles, Program Profiles and Policy Profiles), and produced 14 technical articles for the Department of Energy. Unfortunately, these quantities provide no measure of the value that the team has provided. Only our clients can judge that, and we are confident that they are pleased with the benefits we have offered them.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

DOE OSTI2024

Improving Cost and Efficiency of the Scalable Solid Oxide Fuel Cells Power System

The objective of this project was to design and develop a 20kW range small-scale solid oxide fuel cells (SOFC) power system for applications such as data centers and commercial buildings. The original plan included a 5,000 hours demonstration and a Techno-Economic Analysis (TEA) which were dropped as part of project termination. The original project plan was to use a stack with a cross-flow cell design which had previously been tested for 500 hours at a community college in Malta, NY. However, it was decided to move to the advanced R-SOFC co-flow cell developed under Department of Energy Award DE-FE0031971. The advanced cell design has the advantage of a larger active area for the same manufacturing footprint which results in fewer required cells for the same stack power, hence a higher volumetric power density (kW/L) and lower cost per kW than the original cross-flow cell design. A full SOFC system Simulink model was developed and calibrated with testing data from a fuel cell stack and BOP (balance of plant) components. The simulation results from the calibrated model showed an acceptable match with the experimental data. A structural analysis conducted for various load scenarios indicated no high stress areas for all spatial directions. Major electrical system components were acquired, built and successfully tested. System sensors were verified and validated against controls. Safety checks, a diagnostic check, PID tuning, and control software commissioning tasks were also conducted. The power electronics prototype was delivered and trial testing completed. Balance of Plant component testing and simulation work was conducted to characterize Reformer-Heat Exchanger heat transfer and backpressure and reformer catalyst methane conversion and product selectivity. Simulations were conducted to design the Anode and Cathode fluid passages and size the air-air and fuel-fuel heat exchangers. A Burner operation map was created from test data and the Anode Gas Recirculation blower was tested to evaluate its durability. The SOFC system used a horizontal style design where components sit directly on a casting with a direct connection to the skid. This design has efficient packaging and a small footprint with approximate dimensions of 750 mm x 700 mm x 1700 mm. An SOFC system was built and successfully tested at the Malta, NY facility The system for over 500 hours under load of which over 300 hours was at full load of 20 kW.

30 DIRECT ENERGY CONVERSION↗

DOE OSTI2024

DECOVALEX-2023: Task D Final Report

Task D of DECOVALEX-2023 is focused on the simulation of the coupled thermal hydraulic-mechanical (THM) behaviour in the full-scale engineered barrier system (EBS). The Horonobe EBS experiment is the demonstration of the full-scale EBS in the underground research laboratory (URL) (performed by JAEA in the Horonobe URL in Japan). Task D consisted of the three steps, a preliminary step (Step 0), simulation of the laboratory tests (Step 1) and simulation of the in-situ full-scale EBS experiment (Step 2). Since the Horonobe EBS experiment demonstrates the vertical emplacement option of the EBS, the experiment gallery is also backfilled with the backfill material. Therefore, interaction between the EBS and the backfill material can also be demonstrated, such as deformation (change of density) of the buffer material. The underground water in the Horonobe URL is saline. This fact adds chemical processes to THM behaviour. For example, mechanical properties (such as swelling pressure of the buffer material and backfill material) and hydraulic properties (such as permeability of the buffer material and backfill material) change depending on the water chemistry. Task D was therefore a challenging Task focused on not only the relatively simple THM behaviour but also complex THM behaviour including chemical processes. Six research teams (BGR, CAS, JAEA, KAERI, SNL and Taipower) participated the Task D. BGR, CAS, JAEA, KAERI and Taipower research teams selected a THM approach, while the SNL research team selected a TH approach. Step 1 involved the simulation of laboratory test results and was important to check the numerical codes developed by the research teams. Step 1 was divided into four sub steps. The simulation results through the Step 1 identified the parameters for simulation of the Step 2. Basic parameters of the materials (buffer material, backfill material, rock mass, concrete, sand) were provided by JAEA. Special parameters which research team needed were identified by back analysis of Step 1. Most notably the mechanical behaviour of swelling and displacement depended on the applied model (elastic model or elastoplastic model). Parameters such as Young’s modulus were found to need smaller values than characterised in the fundamental laboratory test results (Step 1-1, 1-2) for the elastic model. Although laboratory experiments are usually simple, test results contained some error. For example, if the saturation level is 100 % or higher, it should be considered an error. This situation was presented in the Step 1-3. A possible reason is that the buffer material is a mixture of bentonite and silica sand. When a specimen is cut to measure volume or weight, sand grains will affect the measurement data. In Step 2, boundary conditions such as temperature on the surface of the simulated overpack, heater power of the electrical heaters installed in the simulated overpack, injection pressure and inflow rate of the test water, were applied. The outer boundary conditions can be selected using measured data (injection pressure and inflow rate of the test water that is controlled by the injection systems installed in the sand layer around the buffer material and in the boundary between backfill material and concrete support). Since such measured data has some noise, research teams developed their own simplified boundary conditions. Inner boundary conditions can be selected using measured data as heater power and temperature on the surface of the simulated overpack. These data also contain some noise, so research teams developed their own simplified developed boundary conditions. Task D validated various approaches thorough the simulation of the in-situ full scale EBS system including backfill of the gallery: variations in the coupling processes (THM or THC), analysis codes, and boundary conditions. Temperature distribution in the buffer material was simulated well by all research teams. This means thermal behaviour is not sensitive to the simulation approaches. Although the water content distribution on the outside of the buffer material was well simulated by all research teams, the simulation results differ from the measured values inside the buffer material (at the centre and inside, near the simulated overpack). The buffer material is made from tap water, but in the in-situ experiment, saline groundwater infiltrates the buffer material. Therefore, the selection of the hydraulic parameters of the buffer material greatly affects the simulation results of the re saturation behaviour of the buffer material. In the Horonobe EBS experiment, measured values suitable for validating the simulation results were not obtained near the simulated overpack. When simulating the pressure and deformation of the buffer material, the measurement data is easily affected by the installation conditions of the measurement sensors, so verifying the measurement data itself remains an issue. Mechanical simulation results differ depending on whether they are considered as elastic or elastoplastic phenomena. The accuracy of measured in-situ data can be assessed by detailed analysis comparing sampling specimen analysis and measured data. The Horonobe EBS experiment is scheduled to be dismantled in the future (FY2026 and 2027). This detailed dismantling investigation will finally confirm the measured data.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2025

Meta‐Analysis and Regression Modeling of the Impacts of Four Indoor Environmental Quality Metrics on Office Performance

Awareness of how buildings interact with the occupant experience—especially human performance—is becoming more prevalent, as seen by increasing interest and investment in healthy built environments. However, there is a need to synthesize the wide array of existing indoor environmental assessment and performance research in a way that can translate directly to building design and operation. Existing research in this area typically focuses on a single isolated metric and has not focused on making the results utilizable by building practitioners. The aim of this research is to investigate existing office performance literature through meta‐analyses and produce regression models for four indoor environmental quality (IEQ) metrics to support critical decision‐making for building operation and renovation. To reach this aim, a literature review was conducted to identify studies that measure the impact of changing ventilation rate, temperature, horizontal illuminance, and noise level in offices on occupant task performance. This repository of field and laboratory studies was analyzed to visualize the trends between the selected IEQ metrics and task performance. The temperature, ventilation rate, and horizontal illuminance regression models showed clear improvement potential when modifying indoor conditions toward the defined high‐performance range, while the regression model for noise level was inconclusive. The discussion notes the importance of designing holistically for all components of these IEQ categories to utilize the results, for example, good filtration on outdoor air for quantifying ventilation impact and uniform overhead lighting with low contrast for quantifying horizontal illuminance impact. The novelty of this work is in considering multiple facets of the indoor environment under a single, unified analysis schema and producing IEQ‐based performance gains that can directly inform cost‐benefit analyses of building design and renovation.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2025

Can Large Language Models Understand Intermediate Representations?

Intermediate Representations (IRs) are essential in compiler design and program analysis, yet their comprehension by Large Language Models (LLMs) remains underexplored. This paper presents a pioneering empirical study to investigate the capabilities of LLMs, including GPT-4, GPT-3, Gemma 2, LLaMA 3.1, and Code Llama, in understanding IRs. We analyze their performance across four tasks: Control Flow Graph (CFG) reconstruction, decompilation, code summarization, and execution reasoning. Our results indicate that while LLMs demonstrate competence in parsing IR syntax and recognizing high-level structures, they struggle with control flow reasoning, execution semantics, and loop handling. Specifically, they often misinterpret branching instructions, omit critical IR operations, and rely on heuristic-based reasoning, leading to errors in CFG reconstruction, IR decompilation, and execution reasoning. The study underscores the necessity for IR-specific enhancements in LLMs, recommending fine-tuning on structured IR datasets and integration of explicit control flow models to augment their comprehension and handling of IR-related tasks.

Jiang, Hailong↗

DOE OSTI2025

AI foundation models for experimental fusion tasks

Artificial Intelligence (AI) foundation models, while successful in various domains of language, speech, and vision, have not been adopted in production for fusion energy experiments. This brief paper presents how AI foundation models can be used for fusion energy diagnostics, enabling, for example, visual automated logbooks to provide greater insights into chains of plasma events in a discharge, in time for between-shot analysis.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2024

Analyzing the impact of design factors on solar module thermomechanical durability using interpretable machine learning techniques

Solar modules in utility-scale systems are expected to maintain decades of lifetime to rival conventional energy sources. However, cyclic thermomechanical loading often degrades their long-term performance, highlighting the importance of effective design to mitigate thermal expansion mismatches between module materials. Given the complex composition of solar modules, isolating the impact of individual components on overall durability remains a challenging task. In this work, we analyze a comprehensive data set that comprises bill-of-materials (BOM) and thermal cycling power loss from 251 distinct module designs to identify the predominant design factors and their impacts on the thermomechanical durability of modules. The methodology of our analysis combines machine learning modeling (random forest) and Shapley additive explanation (SHAP) to correlate design factors with power loss and interpret the model’s decision-making. The interpretation reveals that silicon type (monocrystalline or polycrystalline), encapsulant thickness, busbar numbers, and wafer thickness predominantly influence the degradation. With lower power loss of around 0.6% on average in the SHAP analysis, monocrystalline cells present better durability than polycrystalline cells. This finding is further substantiated by statistical testing on our raw data set. The SHAP analysis also demonstrates that while thicker encapsulants lead to reduced power loss, further increasing their thickness over around 0.6 to 0.7 mm does not yield additional benefits, particularly for the front side one. In addition, other important BOM features such as the number of busbars are analyzed. This study provides a blueprint for utilizing explainable machine learning techniques in a complex material system and can potentially guide future research on optimizing the design of solar modules.

14 SOLAR ENERGY↗

DOE OSTI2026

Dara: Automated Multiple-Hypothesis Phase Identification and Refinement from Powder X-ray Diffraction

Powder X-ray diffraction (XRD) is a foundational technique for characterizing crystalline materials. However, the reliable interpretation of XRD patterns, particularly in multiphase systems, remains a manual and expertise-demanding task. As a characterization method that only provides structural information, multiple reference phases can often be fit to a single pattern, leading to potential misinterpretation when alternative solutions are overlooked. To ease humans’ efforts and address the challenge, we introduce Dara (data-driven automated Rietveld analysis), a framework designed to automate the robust identification and refinement of multiple phases from powder XRD data. Dara performs an exhaustive tree search over all plausible phase combinations within a given chemical space and validates each hypothesis using the BGMN Rietveld refinement routine. Key features include structural database filtering, automatic clustering of isostructural phases during tree expansion, and peak-matching-based scoring to identify promising phases for refinement. When ambiguity exists, Dara generates multiple hypothesis which can then be decided between by human experts or with further characterization tools. By enhancing the reliability and accuracy of phase identification, Dara enables scalable analysis of realistic complex XRD patterns and provides a foundation for integration into multimodal characterization workflows, moving toward fully self-driving materials discovery.

Biological databases↗

DOE OSTI2024

PhysBERT: A text embedding model for physics scientific literature

The specialized language and complex concepts in physics pose significant challenges for information extraction through Natural Language Processing (NLP). Central to effective NLP applications is the text embedding model, which converts text into dense vector representations for efficient information retrieval and semantic analysis. In this work, we introduce PhysBERT, the first physics-specific text embedding model. Pre-trained on a curated corpus of 1.2 × 106 arXiv physics papers and fine-tuned with supervised data, PhysBERT outperforms leading general-purpose models on physics-specific tasks, including the effectiveness in fine-tuning for specific physics subdomains.

Hellert, Thorsten (ORCID:0000000227970926)↗

DOE OSTI2026

Scaling Spectroradiometric Monitoring to an Algal Factory Farm

This report briefly summarizes the spectroradiometric approach adapted and deployed as part of Task 4.2 within the BETO-funded D evelopment of I ntegrated S creening, C ultivar O ptimization, and V erification R esearch (DISCOVR) project, specifically task 4, “Increase crop protection and culture stability.” The goals of this subtask were to monitor ponds cultivating a variety of DISCOVR strains at the Arizona Center for Algae Technology and Innovation (AzCATI) with spectroradiometric technology to: • Predict biomass concentrations with high temporal frequency. • Optimize automated analysis methods that enabled feedback for pond automation strategies • Identify pond crash indicators from primary (pest/predator) and indirect (change in algal host) spectroscopic signals with potential for early treatment/intervention • Provide recommendations for technology deployment at scale. This report fulfills the deliverable requirement for the final bullet “provide recommendations for technology deployment at scale”.

09 BIOMASS FUELS↗

DOE OSTI2026

Optimizing Neutrino Flavor Conversion Measurements through Machine Learning

The phenomenon of neutrino flavor conversion whereby the flavor of a neutrino particle can change between its time of production and later detection was the first definitive evidence of physics beyond the Standard Model. Some of the oscillation parameters used to describe this conversion are not yet well measured, leaving important questions still open regarding flavor conversion both in vacuum and as neutrinos travel through matter. NOvA is a long-baseline neutrino oscillation experiment that uses Fermilab's predominantly $\nu_\mu$ NuMI beam. A 14 kton oil-based liquid scintillator far detector 810 km away is used to measure neutrino oscillation through the $\nu_\mu$ disappearance and $\nu_e$ appearance channels. Super-K is a 50 kton water Cherenkov detector, which measures the disappearance of $\nu_e$ produced during solar fusion. The high density environment of the sun decreases the $\nu_e$ survival probability at higher energies observable in Super-K compared to the vacuum-dominated oscillations at lower energies. However, the transition region is overshadowed by radioactive background in the detector. In both of these experiments, the separation of neutrino detection events from background and classification of neutrino flavor are crucial tasks that benefit from the introduction of machine learning. Chapter 1 gives an overview of neutrinos and the context under which flavor conversion is measured in this dissertation. Chapters 2-7 present the results of a Bayesian sampling approach for the latest NOvA 3-flavor oscillation analysis with $26.6 \times 10^{20}$ protons on target in neutrino mode and $12.5 \times 10^{20}$ in antineutrino mode collected over 10 years. Chapters 8-13 present the results of extending the Super-K solar analysis to lower energies during its fourth phase with 2970 days of livetime.

Yankelevich, Alejandro Jaime [UC, Irvine]↗

DOE OSTI2026

Optimizing Neutrino Flavor Conversion Measurements through Machine Learning

The phenomenon of neutrino flavor conversion whereby the flavor of a neutrino particle can change between its time of production and later detection was the first definitive evidence of physics beyond the Standard Model. Some of the oscillation parameters used to describe this conversion are not yet well measured, leaving important questions still open regarding flavor conversion both in vacuum and as neutrinos travel through matter. NOvA is a long-baseline neutrino oscillation experiment that uses Fermilab's predominantly $\nu_\mu$ NuMI beam. A 14 kton oil-based liquid scintillator far detector 810 km away is used to measure neutrino oscillation through the $\nu_\mu$ disappearance and $\nu_e$ appearance channels. Super-K is a 50 kton water Cherenkov detector, which measures the disappearance of $\nu_e$ produced during solar fusion. The high density environment of the sun decreases the $\nu_e$ survival probability at higher energies observable in Super-K compared to the vacuum-dominated oscillations at lower energies. However, the transition region is overshadowed by radioactive background in the detector. In both of these experiments, the separation of neutrino detection events from background and classification of neutrino flavor are crucial tasks that benefit from the introduction of machine learning. Chapter 1 gives an overview of neutrinos and the context under which flavor conversion is measured in this dissertation. Chapters 2-7 present the results of a Bayesian sampling approach for the latest NOvA 3-flavor oscillation analysis with $26.6 \times 10^{20}$ protons on target in neutrino mode and $12.5 \times 10^{20}$ in antineutrino mode collected over 10 years. Chapters 8-13 present the results of extending the Super-K solar analysis to lower energies during its fourth phase with 2970 days of livetime.

Yankelevich, Alejandro Jaime [UC, Irvine]↗

DOE OSTI2026

Optimizing Neutrino Flavor Conversion Measurements through Machine Learning

The phenomenon of neutrino flavor conversion whereby the flavor of a neutrino particle can change between its time of production and later detection was the first definitive evidence of physics beyond the Standard Model. Some of the oscillation parameters used to describe this conversion are not yet well measured, leaving important questions still open regarding flavor conversion both in vacuum and as neutrinos travel through matter. NOvA is a long-baseline neutrino oscillation experiment that uses Fermilab's predominantly $\nu_\mu$ NuMI beam. A 14 kton oil-based liquid scintillator far detector 810 km away is used to measure neutrino oscillation through the $\nu_\mu$ disappearance and $\nu_e$ appearance channels. Super-K is a 50 kton water Cherenkov detector, which measures the disappearance of $\nu_e$ produced during solar fusion. The high density environment of the sun decreases the $\nu_e$ survival probability at higher energies observable in Super-K compared to the vacuum-dominated oscillations at lower energies. However, the transition region is overshadowed by radioactive background in the detector. In both of these experiments, the separation of neutrino detection events from background and classification of neutrino flavor are crucial tasks that benefit from the introduction of machine learning. Chapter 1 gives an overview of neutrinos and the context under which flavor conversion is measured in this dissertation. Chapters 2-7 present the results of a Bayesian sampling approach for the latest NOvA 3-flavor oscillation analysis with $26.6 \times 10^{20}$ protons on target in neutrino mode and $12.5 \times 10^{20}$ in antineutrino mode collected over 10 years. Chapters 8-13 present the results of extending the Super-K solar analysis to lower energies during its fourth phase with 2970 days of livetime.

Yankelevich, Alejandro Jaime [UC, Irvine] (ORCID:0↗