Search NASA⌕ Search

SEARCH · Search NASA

Results for “Operator learning”

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 415 records · Page 23
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Improving the Freight Productivity of a Heavy-Duty, Battery Electric Truck by Intelligent Energy Management

This project aimed to enhance the range and reduce the operating costs of battery electric Class 8 trucks traveling over 250 miles daily. This was achieved through the development and implementation of an intelligent-Energy Management System (i-EMS) that leverages vehicle and operations data, physics-aware machine learning algorithms, and vehicle-to-cloud (V2C) connectivity. The project hypothesized that advanced machine learning algorithms and real-time data analytics could significantly improve the energy efficiency and range of these trucks. Key objectives included developing a physics-aware machine learning algorithm, implementing an i-EMS with V2C connectivity and physics-aware spatial data analytics (PSDA), and validating the system’s effectiveness with fleet partners HEB Companies and Murphy Logistics. Extensive data collection from vehicle operations, including vehicle characteristics, road conditions, and payload, was conducted. A machine learning algorithm was developed to predict energy consumption and enable proactive decision-making. The i-EMS was implemented on two Volvo VNR BEVs, with operators receiving charging and routing recommendations. Charging stations were installed at depot locations in Texas and Minnesota, with an additional on-route charger in Minnesota. Significant findings included a 14% range improvement for Murphy Logistics on a highway-driving eco-route and a 22% range improvement for HEB Companies on a city-driving eco-route. The i-EMS utilized rule-based methods and physics-based algorithms to predict and reduce energy consumption, with real-time monitoring and analysis through V2C connectivity enabling proactive decision-making. The project demonstrated the feasibility and economic viability of battery electric Class 8 trucks for long-haul operations, showcasing the potential of physics-aware machine learning in optimizing energy management. The successful implementation of the i-EMS in real-world scenarios validates its practical application and effectiveness, paving the way for the widespread adoption of battery electric vehicles in the freight transportation industry.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Machine learning for reactor power monitoring with limited labeled data

Real-time reactor power monitoring is critical for a variety of nuclear applications, spanning safety, security, operations, and maintenance. While machine learning methods have shown promise in monitoring reactor power levels, there is limited research on their efficacy in label-starved environments. The goal of this work is to assess the feasibility of classifying nuclear reactor power level using multisource data in scenarios with limited labels. Data were collected using low-resolution multisensors at four nuclear reactor facilities: two large research reactors and two TRIGA reactors. Within each pair, one reactor dataset served as the source and the other as the target in a transfer learning paradigm. Twenty-three supervised models were trained on labeled sequences of magnetic field and acceleration data from each of the target sites. Self-learning and transfer learning methods were applied to the top performing models to assess their classification performance with increasing amounts of labeled data. While reactor power level classification was achieved with a Matthews Correlation Coefficient of up to 0.739 ± 0.003 and 0.622 ± 0.009 with only 400 sequences per power state for the large research reactor and TRIGA target sites, respectively, self-learning and transfer learning leveraging source site data did not improve target classification performance. These findings suggest that alternative methods, such as higher sensitivity sensors, digital twins, or the use of physics-informed models, are required to enable high-performance classification in machine learning approaches to reactor monitoring with a dearth of target ground truth.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

Atikokan Digital Twin: Machine learning in a biomass energy system

The Atikokan Generating Station, operated by Ontario Power Generation, has a 200 MW, biomass-fired tower boiler that operates on a dispatch schedule with a five-minute cycle. The boiler is generally operated in the range of 40–100 MW using two of five burner levels. In order to optimize boiler performance, we propose the implementation of a unique digital twin. Our digital twin abstraction couples Bayesian inference from science-based models and from observations (machine learning) with decision theory to predict operating-variable set points that optimize the physical asset (the boiler) in the presence of uncertainty (artificial intelligence). We focus this paper on the continuous Bayesian machine learning part of the Atikokan Digital Twin; we discuss decision theory in a companion paper. We identify and learn about 12 operational, model, and measured-output parameters and their uncertainties from high-fidelity, science-based simulations of the Atikokan boiler and from the observed measurements at the power plant. Since the goal of the Atikokan Digital Twin is to implement it online in real time, we require fast function evaluations for the quantities of interest extracted from the simulations in the Bayesian analysis. We use Gaussian process regression/interpolation to create accurate, robust surrogate models. We define the Bayesian priors and likelihood function and solve for the posterior distributions of the 12 parameters. Here we then propagate these distributions (i.e., parameters with uncertainty) into the predicted distributions of 790 quantities of interest to learn about the relative importance of various sources of error including experimental, model, and operating-parameter errors.

09 BIOMASS FUELS↗
DOE OSTI2025

Self-adaptive weights based on balanced residual decay rate for physics-informed neural networks and deep operator networks

Physics-informed deep learning has emerged as a promising alternative for solving partial differential equations. However, for complex problems, training these networks can still be challenging, often resulting in unsatisfactory accuracy and efficiency. In this work, we demonstrate that the failure of plain physics-informed neural networks arises from the significant discrepancy in the convergence rate of residuals at different training points, where the slowest convergence rate dominates the overall solution convergence. Based on these observations, we propose a pointwise adaptive weighting method that balances the residual decay rate across different training points. The performance of our proposed adaptive weighting method is compared with current state-of-the-art adaptive weighting methods on benchmark problems for both physics-informed neural networks and physics-informed deep operator networks. In conclusion, through extensive numerical results we demonstrate that our proposed approach of balanced residual decay rates offers several advantages, including bounded weights, high prediction accuracy, fast convergence rate, low training uncertainty, low computational cost, and ease of hyperparameter tuning.

Balanced convergence rate↗
DOE OSTI2024

Deep learning based event reconstruction for cyclotron radiation emission spectroscopy

The objective of the cyclotron radiation emission spectroscopy (CRES) technology is to build precise particle energy spectra. This is achieved by identifying the start frequencies of charged particle trajectories which, when exposed to an external magnetic field, leave semi-linear profiles (called tracks) in the time–frequency plane. Due to the need for excellent instrumental energy resolution in application, highly efficient and accurate track reconstruction methods are desired. Deep learning convolutional neural networks (CNNs) - particularly suited to deal with information-sparse data and which offer precise foreground localization—may be utilized to extract track properties from measured CRES signals (called events) with relative computational ease. In this work, we develop a novel machine learning based model which operates a CNN and a support vector machine in tandem to perform this reconstruction. A primary application of our method is shown on simulated CRES signals which mimic those of the Project 8 experiment—a novel effort to extract the unknown absolute neutrino mass value from a precise measurement of tritium β - -decay energy spectrum. When compared to a point-clustering based technique used as a baseline, we show a relative gain of 24.1% in event reconstruction efficiency and comparable performance in accuracy of track parameter reconstruction.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2017

Lignite-plus-Biomass to Synthetic Jet Fuel with CO2 Capture and Storage: Design, Cost, and Greenhouse Gas Emissions Analysis for a Near-Term First-of-a-Kind Demonstration Project and Prospective Future Commercial Plants

We report on a 30-month design study for a first-of-a-kind (FOAK) demonstration plant that would be built at a site near Meridian, Mississippi, to coprocess lignite coal and woody biomass into jet fuel. The design uses an oxygen-blown TRIG™ gasifier developed by KBR and Southern Company. Fischer-Tropsch conversion of the syngas produces synthetic paraffinic kerosene (SPK) as the primary product, plus naphtha. Other co-products include electricity sold to the grid and CO 2 sold for use in enhanced oil recovery (EOR). Previous studies have identified coprocessing of various coals and biomass with CO 2 capture as promising options for cost-competitive production of low net lifecycle greenhouse gas (GHG) emissions synthetic fuels. The effort reported here goes beyond earlier studies in the level of detail in process design and cost estimation with the aim of improving the understanding of the economic prospects for lignite and woody biomass coprocessing systems. Key objectives in the design of the FOAK lignite/biomass-to-jet (LBJ) plant were to achieve net lifecycle GHG emissions for the SPK that are less than for conventional petroleum-derived jet fuel and to make process design and equipment selections such that the plant could be built and operated in the near term, e.g., before 2025. The process design was developed by researchers at Princeton University and the University of Queensland and validated by engineers at the WorleyParsons Group (WP). Commercial vendors provided design and cost estimates for several major plant components. Bare-erected capital costs and operating and maintenance costs were estimated by WP. The primary objective in building and operating the FOAK plant would be to demonstrate the technical viability of the LBJ concept as an essential first step toward launching commercial-scale plants in the longer term. With this in mind, the project team developed a set of principles to help guide the process design: the total plant cost should be less than 2 billion USD (to limit investment risk); the level of input biomass should be both proportionally significant to reflect biomass/lignite co-gasification and sufficient to achieve GHG emissions goals; process design decisions and vendor/equipment selections should prioritize the likelihood of technical success over minimizing the cost of jet fuel production. The resulting FOAK plant design capacity is 1,551 metric t/d (45.5% moisture) lignite and 556 t/d (43.3% moisture) biomass, for a total input of 295 MW HHV , of which 25% is biomass. The design output is 1,252 actual barrels per day of liquids, of which 80% is SPK (62.3 MW LHV ) and 20% is naphtha (13.9 MW LHV ). It exports 15 MW e of electricity after satisfying a 38 MW e onsite auxiliary load. Additional products are 1,326 t/d of pressurized pipeline-quality CO 2 and 49 t/day of sulfuric acid (93 wt% H 2 SO 4 ). With thinnings from sustainably-managed southern pine plantations as the biomass, and with captured CO 2 stored underground via EOR, the net lifecycle emissions for the SPK product are estimated to be about one-quarter of those for petroleum-derived jet fuel. The bare-erected cost (BEC) estimated by WP for this plant is 588 million USD (2015). The authors’ best estimate of total plant cost (TPC) is 1,230 million USD, arrived at by assuming engineering, procurement and construction management services (20% of BEC), process 5 contingencies (35% of BEC), and project contingencies (35% of the sum of all other costs, i.e., 35% of 1.55xBEC). Not surprisingly, an annual discounted cash flow (DCF) analysis determined that it would be impossible to generate a positive net present value (NPV) over a wide range in key input assumptions. An SPK production cost subsidy of nearly 400/bbl USD over a 20-year plant life would be required to achieve zero NPV under a baseline set of assumptions (including 3% real weighted average cost of capital and a levelized crude oil price of 80/bbl USD). Alternatively, a capital grant in excess of the TPC value would also achieve zero NPV. The poor financial results reflect the small scale of the plant, the design principle to prioritize technical success, the levels of contingencies appropriate for the relatively early stage of project development, and the first-of-a-kind nature of the plant. Technology innovations, learning via construction and operating experience, alternative plant configurations, and larger scale should improve economics of future plants. To help understand these prospects, a preliminary analysis of N th -of-a-kind (NOAK) plants was developed, but with the limitation that plants would use only equipment components that for the most part are already commercial today: consideration of advanced, not-yet-commercial technologies and of R&D-driven improvements in existing technologies were beyond the scope of this analysis. The analysis found that a variety of NOAK plant designs that coprocess lignite and woody biomass to make jet fuel are unlikely to be economically competitive without subsidy even in the presence of a high future carbon tax or equivalent greenhouse gas mitigation policy. This conclusion applies to process configurations and input biomass/lignite ratios that result in net GHG emissions as high as those for petroleum-derived jet fuel and as low as zero. In contrast, encouraging results were found for plants processing only biomass. The economics of these “BECCS” plants (biomass energy with CCS) improve dramatically with the strength of carbon mitigation policies because of their strongly negative net GHG emissions. These findings do not imply that coal/biomass coprocessing strategies for making synfuels with CCS are not economically promising – only that, in the case of lignite, much more than a 25% biomass coprocessing rate would be needed. However, our analysis shows that all such systems are unlikely to be economic in the absence of a strong carbon mitigation policy. Future R&D driven technological innovations could modify this conclusion. Among other R&D priorities, an emphasis on better understanding and reducing plant auxiliary loads is warranted.

09 BIOMASS FUELS↗
DOE OSTI2025

Artificial Intelligence for Data Center Operations (AIOps): Cooperative Research and Development (Final Report)

High performance computing data centers will increasingly need to rely on automation to keep pace with exascale growth in compute capability and to manage and optimize the data center environment and facility resources. Artificial intelligence and machine learning approaches provide the means to improve HPC data center operational efficiency, by learning historical trends and training models to operate on real-time data collected from both IT and facilities sources. NREL has developed methods of real-time collection, aggregation and streaming of these data in the ESIF HPC Data Center and has collected a significant dataset of relevant metrics across computer systems, racks, environmental, building and utility sources for research into various predictive analytics problems. HPE's Advanced Technology Group (ATG) is doing comprehensive research into exascale monitoring and management for High Performance Computing (HPC) systems (hereinafter HPE's Data Monitoring/ Management Technology). NREL and HPE will collaborate to add Artificial Intelligence (AI) to NREL's real-time data collection/ aggregation/ streaming system and HPE's Data Monitoring/ Management System, with the goal of improving the operational efficiency of NREL's Energy Systems Integration Facility (ESIF) HPC Data Center through data analytics on both historical and real-time data from IT systems and facilities operations. This collaboration will consist of efforts in Data Management, Data Analytics, and AI/ML Optimization for both manual and autonomous intervention in data center operations. This will be a multi-year, multi-staged effort with a goal towards building capabilities for an Advanced Smart Facility, and demonstration of these techniques in the NREL ESIF HPC Data Center.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Robustness of Deep Learning Classification to Adversarial Input on GPUs: Asynchronous Parallel Accumulation Is a Source of Vulnerability

The ability of machine learning (ML) classification models to resist small, targeted input perturbations—known as adversarial attacks—is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets used in robustness assessments. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.

Shanmugavelu, Sanjif [Maxeler Technologies, a Groq↗
DOE OSTI2022

Operation-adversarial scenario generation

This paper proposes a modified conditional generative adversarial network (cGAN) model to generate net load scenarios for power systems that are statistically credible, conditioned by given labels (e.g., seasons), and, at the same time, “stressful” to the system operations and dispatch decisions. The measure of stress used in this paper is based on the operating cost increases due to net load changes. The proposed operation-adversarial cGAN (OA-cGAN) internalizes a DC optimal power flow model and seeks to maximize the operating cost and achieve a worst-case data generation. The training and testing stages employed in the proposed OA-cGAN use historical day-ahead net load forecast errors and has been implemented for the realistic NYISO 11-zone system. In conclusion, our numerical experiments demonstrate that the generated operation-adversarial forecast errors lead to more cost-effective and reliable dispatch decisions.

42 ENGINEERING↗
DOE OSTI2024

The importance of cycle-by-cycle data in performing rapid battery technology development and validation

Lithium-ion battery (LiB) technology is playing a crucial role in transforming the predominantly fossil fuel-based transportation and stationary storage sectors to achieve a low-carbon economy. Rapid innovation in the LiB materials to electrode to cell design is happening to satisfy the performance, life, and safety metrics required by those myriads of applications. Lately, advanced analytics, such as machine-learning or artificial intelligence (ML/AI) techniques, are being used more frequently to aid in expedited LiB technology development, performance validation, and life prediction. The success of these techniques often relies on a large volume of well-defined and high-quality battery test data. On the other hand, most battery developers and research and development (R&D) communities are still following a classical approach to develop batteries, which is running calendar- and/or cycle-aging tests, performing reference performance tests (RPTs), and conducting post-mortem analyses periodically without paying attention to the wealth of data often not collected during the calendar or cycle life aging tests. This sparse data collection approach is time- and resource-intensive, requiring data capture and evaluation of months to years of RPT data to diagnose accurate battery state of performance, health, and safety. Even so, the underlying aging modes and mechanisms can be missed. If collected properly, battery test data during cycling or calendaring can be efficiently combined with ML/AI techniques to create powerful tools in the rapid diagnosis of battery state of performance, health, and safety along with insights into underlying aging modes and mechanisms. In this report, we discuss the importance of effective cycle-by-cycle (CBC) data collection with example case studies. Within a reasonable timeframe, RPT data are often inadequate in capturing many of the crucial battery aging dynamics, which often predominantly show up in CBC test data. Finally, we also show examples of ML/AI techniques that use CBC data in rapid diagnosis and projection of LiB state of health (SOH) to motivate the scientific community in collecting and using CBC data to facilitate expeditious technology development and validation.

25 ENERGY STORAGE↗
DOE OSTI2023

Differentiable physics-enabled closure modeling for Burgers’ turbulence

Abstract Data-driven turbulence modeling is experiencing a surge in interest following algorithmic and hardware developments in the data sciences. We discuss an approach using the differentiable physics paradigm that combines known physics with machine learning to develop closure models for Burgers’ turbulence. We consider the one-dimensional Burgers system as a prototypical test problem for modeling the unresolved terms in advection-dominated turbulence problems. We train a series of models that incorporate varying degrees of physical assumptions on an a posteriori loss function to test the efficacy of models across a range of system parameters, including viscosity, time, and grid resolution. We find that constraining models with inductive biases in the form of partial differential equations that contain known physics or existing closure approaches produces highly data-efficient, accurate, and generalizable models, outperforming state-of-the-art baselines. Addition of structure in the form of physics information also brings a level of interpretability to the models, potentially offering a stepping stone to the future of closure modeling.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Machine learning of hidden variables in multiscale fluid simulation

Solving fluid dynamics equations often requires the use of closure relations that account for missing microphysics. For example, when solving equations related to fluid dynamics for systems with a large Reynolds number, sub-grid effects become important and a turbulence closure is required, and in systems with a large Knudsen number, kinetic effects become important and a kinetic closure is required. By adding an equation governing the growth and transport of the quantity requiring the closure relation, it becomes possible to capture microphysics through the introduction of ‘hidden variables’ that are non-local in space and time. The behavior of the ‘hidden variables’ in response to the fluid conditions can be learned from a higher fidelity or ab-initio model that contains all the microphysics. In our study, a partial differential equation simulator that is end-to-end differentiable is used to train judiciously placed neural networks against ground-truth simulations. We show that this method enables an Euler equation based approach to reproduce non-linear, large Knudsen number plasma physics that can otherwise only be modeled using Boltzmann-like equation simulators such as Vlasov or particle-in-cell modeling.

42 ENGINEERING↗
DOE OSTI2018

Production Facility Prototype Blower 1000 Hour Test Results III

Long duration operations tests of the Aerzen GM 12.4 roots style blower in a closed loop configuration provides valuable data and lessons learned for long-term operation at the Mo-99 production facility. The blower was operated in a closed loop configuration with the flow conditions anticipated in plant operation with a Mo-100 target inline. The additional thermal energy generated from beam heating of the Mo-100 disks was not included in these tests. Eight long duration 500 and 1000 hour tests have been completed since the first test was performed in January of 2016. All eight long duration tests have proven successful in exposing preventable issues related to oil and helium leaks.

47 OTHER INSTRUMENTATION↗
DOE OSTI2018

Spectral Methods in Time-dependent Data Analysis

We aim to create a new model for time-dependent data analysis, named dynamical learning, that integrates data-driven manifold learning techniques with operator-theoretic methods from dynamical systems theory. This approach has the potential to deliver more efficient methods for analyzing time-dependent data, such as video streams, by naturally separating out the temporal and spatial features of the data. We aim to apply the newly developed methods to video surveillance data related to Sandia mission applications, and particularly focus on the problems of image segmentation and object tracking. This project ended early due to the departure of the PI from Sandia about 18 months into the project. Therefore, this document reports on partial progress towards the initial goals of the project. In addition, this document reports on part of the work conducted during the project; see the Appendix for a summary of all the work conducted during the 18 months.

97 MATHEMATICS AND COMPUTING↗