Search NASA⌕ Search

SEARCH · Search NASA

Results for “successive training”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Compute in‐Memory with Non‐Volatile Elements for Neural Networks: A Review from a Co‐Design Perspective

Abstract Deep learning has become ubiquitous, touching daily lives across the globe. Today, traditional computer architectures are stressed to their limits in efficiently executing the growing complexity of data and models. Compute‐in‐memory (CIM) can potentially play an important role in developing efficient hardware solutions that reduce data movement from compute‐unit to memory, known as the von Neumann bottleneck. At its heart is a cross‐bar architecture with nodal non‐volatile‐memory elements that performs an analog multiply‐and‐accumulate operation, enabling the matrix‐vector‐multiplications repeatedly used in all neural network workloads. The memory materials can significantly influence final system‐level characteristics and chip performance, including speed, power, and classification accuracy. With an over‐arching co‐design viewpoint, this review assesses the use of cross‐bar based CIM for neural networks, connecting the material properties and the associated design constraints and demands to application, architecture, and performance. Both digital and analog memory are considered, assessing the status for training and inference, and providing metrics for the collective set of properties non‐volatile memory materials will need to demonstrate for a successful CIM technology.

36 MATERIALS SCIENCE↗
DOE OSTI2022

A learned closure method applied to phase mixing in a turbulent gradient-driven gyrokinetic system in simple geometry

We present a new method for formulating closures that learn from kinetic simulation data. We apply this method to phase mixing in a simple gyrokinetic turbulent system – temperature-gradient-driven turbulence in an unsheared slab. The closure, called the learned multi-mode (LMM) closure, is constructed by, first, extracting an optimal basis from a nonlinear kinetic simulation using singular value decomposition. Subsequent nonlinear fluid simulations are projected onto this basis and the results are used to formulate the closure. We compare the closure with other closures schemes over a broad range of the relevant two-dimensional parameter space (collisionality and gradient drive). We find that the turbulent kinetic system produces phase-mixing rates much lower than the linear expectations, which the LMM closure is capable of capturing. We also compare radial heat fluxes. A Hammett–Perkins closure, generalized to include collisional effects, is quite successful throughout the parameter space, producing ${\sim }14\,\%$ root-mean-square (r.m.s.) error. The LMM closure is also very effective: when trained at three (two) points (in a 35 point parameter grid), the LMM closure produces $8\,\%$ ( $12\,\%$ ) r.m.s. errors. The LMM procedure can be readily generalized to other closure problems.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2026

Regulation compliant AI for fusion: explainable image-based feedback control of divertor detachment in DIII-D tokamak

While artificial intelligence (AI) has been promising for fusion control, its inherent black-box nature will make compliant implementation in regulatory environments a challenge. This study implements and validates a real-time AI-enabled linear and interpretable control system for successful divertor detachment control with the DIII-D lower divertor camera. Using D 2 gas, we demonstrate successful feedback divertor detachment control with a mean absolute difference of 2% from the target for both detachment and reattachment. This automatic training and linear processing framework can be extended to any image-based diagnostic for future fusion reactors.

computer vision↗
DOE OSTI2022

Deep reinforcement learning with online data augmentation to improve sample efficiency for intelligent HVAC control

Deep Reinforcement Learning (DRL) has started showing success in real-world applications such as building energy optimization. Much of the research in this space utilized simulated environments to train RL-agent in an offline mode. Very few research have used DRL-based control in real-world systems due to two main reasons: 1) sample efficiency challenge---DRL approaches need to perform a lot of interactions with the environment to collect sufficient experiences to learn from, which is difficult in real systems, and 2) comfort or safety related constraints---user's comfort must never or at least rarely be violated. In this work, we propose a novel deep Reinforcement Learning framework with online Data Augmentation (RLDA) to address the sample efficiency challenge of real-world RL. We used a time series Generative Adversarial Network (TimeGAN) architecture as a data generator. We further evaluated the proposed RLDA framework using a case study of an intelligent HVAC control. With a ≈28% improvement in the sample efficiency, RLDA framework lays the way towards increased adoption of DRL-based intelligent control in real-world building energy management systems.

Kurte, Kuldeep↗
DOE OSTI2022

Optimization of Water-Alternating-CO2 Injection Field Operations Using a Machine-Learning-Assisted Workflow

Summary This paper will present a robust workflow to address multiobjective optimization (MOO) of carbon dioxide (CO2)-enhanced oil recovery (EOR)-sequestration projects with a large number of operational control parameters. Farnsworth unit (FWU) field, a mature oil reservoir undergoing CO2 alternating water injection (CO2-WAG) EOR, will be used as a field case to validate the proposed optimization protocol. The expected outcome of this work would be a repository of Pareto-optimal solutions of multiple objective functions, including oil recovery, carbon storage volume, and project economics. FWU’s numerical model is used to demonstrate the proposed optimization workflow. Because using MOO requires computationally intensive procedures, machine-learning-based proxies are introduced to substitute for the high-fidelity model, thus reducing the total computation overhead. The vector machine regression combined with the Gaussian kernel (Gaussian-SVR) is used to construct proxies. An iterative self-adjusting process prepares the training knowledge base to develop robust proxies and minimizes computational time. The proxies’ hyperparameters will be optimally designed using Bayesian optimization to achieve better generalization performance. Trained proxies will be coupled with multiobjective particle swarm Optimization (MOPSO) protocol to construct the Pareto-front solution repository. The outcomes of this workflow will be a repository containing Pareto-optimal solutions of multiple objectives considered in the CO2-WAG project. The proposed optimization workflow will be compared with another established methodology using a multilayer neural network (MLNN) to validate its feasibility in handling MOO with a large number of parameters to control. Optimization parameters used include operational variables that might be used to control the CO2-WAG process, such as the duration of the water/gas injection period, producer bottomhole pressure (BHP) control, and water injection rate of each well included in the numerical model. It is proved that the workflow coupling Gaussian-SVR proxies and the iterative self-adjusting protocol is more computationally efficient. The MOO process is made more rapid by squeezing the size of the required training knowledge base while maintaining the high accuracy of the optimized results. The outcomes of the optimization study show promising results in successfully establishing the solution repository considering multiple objective functions. Results are also verified by validating the Pareto fronts with simulation results using obtained optimized control parameters. The outcome from this work could provide field operators an opportunity to design a CO2-WAG project using as many inputs as possible from the reservoir models. The proposed work introduces a novel concept that couples Gaussian-SVR proxies with a self-adjusting protocol to increase the computational efficiency of the proposed workflow and to guarantee the high accuracy of the obtained optimized results. More importantly, the workflow can optimize a large number of control parameters used in a complex CO2-WAG process, which greatly extends its utility in solving large-scale MOO problems in various projects with similar desired outcomes.

Energy & Fuels↗
DOE OSTI2023

Study of The Compression of an Imploding Plasma and the Magnetic- Field Flux Using Advanced, Non-Intrusive, Spectroscopic Techniques

The main motivation of the project was to measure the magnetic field in pulsed power driven plasma compression using advanced non-intrusive spectroscopic techniques. The strength of the project is that it encompasses a combination of laboratory experiments with diagnostics for both the magnetic field and plasma properties and computational modeling. The following issues were addressed: (1) Plasma and magnetic field dynamics, (2) Field compression and diffusion, (3) Energy balance in the system, (4) Onset of instabilities, and (5) Correlation of the plasma and field evolutions with the initial conditions. The unique feature of the research was the use of advanced spectroscopy to measure both the magnetic field and various plasma properties in multiple locations using both side-on and end-on views, simultaneously. The experimental system allowed for imploding different gas mixtures (which facilitated local measurements) and for flexibility in modifying the initial conditions. Two current generators were used with significantly different current rise times (200 ns and 1.6 µs) but at the same current level of 400 kA. The slower pulsed power generator (at rise time 1.6 µs) was based at the Weizmann Institute of Science (WIS) and imploded plasma into which an axial magnetic field (up to 1 T) was initially embedded. The faster current rise time generator (at rise time 200 ns) based at University of California San Diego (UCSD) was used for plasma parameter studies initially, with collaborative efforts we successfully implemented the advanced spectroscopy techniques at UC San Diego. One graduate student (Nick Aybar) and a postdoctoral scholar (Maylis Dozeries) were trained. They both are based at the Lawrence Livermore National Laboratory now.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

MedNAS: Multiscale Training-Free Neural Architecture Search for Medical Image Analysis

Deep neural networks have demonstrated impressive results in medical image analysis, but designing suitable architectures for each specific task is expertise dependent and time consuming. Neural architecture search (NAS) offers an effective means of discovering architectures. It has been highly successful in numerous applications, particularly in natural image classification. Yet, medical images possess unique characteristics, such as small regions and a wide variety of lesion sizes, that differentiate them from natural images. Furthermore, most current NAS methods struggle with high computational costs, especially when dealing with high-resolution image datasets. In this article, we present a novel evolutionary NAS method called multiscale training-free neural architecture search (MSTF-NAS) to address these challenges. Specifically, to accommodate the broad range of lesion region sizes in disease diagnosis, we develop a new reduction cell search space that enables the search algorithm to explicitly identify the optimal scale combination for multiscale feature extraction. Further, to overcome the issue of high computational costs, we utilize training-free indicators as performance measures for candidate architectures, which allows us to search for the optimal architecture more efficiently. More specifically, by considering the capability and simplicity of various networks, we formulate a multiobjective optimization problem that involves two training-free indicators and model complexity for candidate architectures. Extensive experiments on a large medical image benchmark and a publicly available breast cancer detection dataset are conducted. The empirical results demonstrate that our MSTF-NAS outperforms both human-designed architectures and current state-of-the-art NAS algorithms on both datasets, indicating the effectiveness of our proposed method.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2020

Machine learning control for disruption and tearing mode avoidance

In this work, real-time feedback control based on machine learning algorithms (MLA) was successfully developed and tested on DIII-D plasmas to avoid tearing modes and disruptions while maximizing the plasma performance, which is measured by normalized plasma beta. Control uses MLAs that were trained with ensemble learning methods using only the data available to the real-time Plasma Control System (PCS) from several thousand DIII-D discharges. A `tearability' metric that quantifies the likelihood of the onset of 2/1 tearing modes (TM) in a given time window, and a 'disruptivity' metric that quantifies the likelihood of the onset of plasma disruptions were first tested o -line then implemented on the PCS. A real-time control system based on these MLAs was successfully tested on DIII-D discharges, using feedback algorithms to maximize β Ν while avoiding tearing modes and to dynamically adjust ramp down to avoid high-current disruptions in ramp down.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2023

A new method for predicting hurricane rapid intensification based on co-occurring environmental parameters

Abstract Tropical cyclones (TCs) that undergo Rapid Intensification (RI) can pose serious socioeconomic threats and can potentially result in major damaging impacts along coastal areas. Considering the complexity of various physical mechanisms that play a role in RI and its relatively low probability of occurrence, predicting RI remains a major operational challenge. In this study, we propose a simple deterministic binary classification model based on the co-occurrence of environmental parameters (MCE) to predict an RI event. More specifically, the model determines the possibility of RI based on a simple count of the number of environmental predictors deemed favorable and unfavorable. We compare our model results to logistic regression (LR) and decision tree (DT) models, well-trained using the same set of environmental predictors. Results reveal that at an RI threshold of 30 kt, the MCE exhibits a critical success index score of 0.233 which is 14% higher than DT and LR model performances. When tested at multiple RI thresholds, the MCE displays relatively higher skill scores across multiple metrics. By simultaneously evaluating the favorability of predictors, the MCE is able to comparatively reduce the number of false alarms predicted when certain predictors are unfavorable toward RI. Interpreting these model results to gain a physical understanding of how co-occurring environmental parameters can affect RI, we highlight future directions for using models based on the MCE approach to understand and predict TC RI as well as other meteorological extremes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

L-PBF High-Throughput Data Pipeline Approach for Multi-modal Integration

Abstract Metal-based additive manufacturing requires active monitoring solutions for assessing part quality. Multiple sensors and data streams, however, generate large heterogeneous data sets that are impractical for manual assessment and characterization. In this work, an automated pipeline is developed that enables feature extraction from high-speed camera video and multi-modal data analysis. The framework removes the need for manual assessment through the utilization of deep learning techniques and training models in a weakly supervised paradigm. We demonstrate this pipeline’s capability over 700,000 high-speed camera frames. The pipeline successfully extracts melt pool and spatter geometries and links them to corresponding pyrometry, radiography, and processparameter information. 715 individual prints are examined to reveal melt pool areas that exceeds 0.07 mm 2 and pyrometry signal over a threshold (375 pyrometry units) were more likely to have defects. These automated processes enable massive throughput of characterization techniques.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Emerging opportunities for hybrid perovskite solar cells using machine learning

While there are several bottlenecks in hybrid organic–inorganic perovskite (HOIP) solar cell production steps, including composition screening, fabrication, material stability, and device performance, machine learning approaches have begun to tackle each of these issues in recent years. Different algorithms have successfully been adopted to solve the unique problems at each step of HOIP development. Specifically, high-throughput experimentation produces vast amount of training data required to effectively implement machine learning methods. Here, we present an overview of machine learning models, including linear regression, neural networks, deep learning, and statistical forecasting. Experimental examples from the literature, where machine learning is applied to HOIP composition screening, thin film fabrication, thin film characterization, and full device testing, are discussed. These paradigms give insights into the future of HOIP solar cell research. As databases expand and computational power improves, increasingly accurate predictions of the HOIP behavior are becoming possible.

Hering, Abigail R. (ORCID:0000000270806953)↗
DOE OSTI2025

Open Power System Datasets and Open Simulation Engines: A Survey Toward Machine Learning Applications

A major factor behind the success of machine learning (ML) models in multiple domains is the availability and accessibility of large, labeled, and well-organized datasets for training and benchmarking. In comparison, power grid datasets face three major challenges: (i) real-world data is often restricted by regulatory constraints, privacy reasons, or security concerns, making it difficult to obtain and work with; (ii) synthetic datasets, which are created to address these limitations, often have incomplete information and are released using specialized tools, making them inaccessible to the broader community; and, (iii) input-output datasets are difficult to generate through simulation for non-experts because open-source simulators are not known outside the power system community. This survey addresses these challenges by serving as an entry point to publicly available datasets and simulators for researchers venturing in this area. We review the current landscape of open-source power network data, machine models, consumer demand profiles, renewable generation data, and inverter models. We also examine open-source power system simulators, which are crucial for generating high-quality, high-fidelity power grid datasets. We aim to provide a foundation for overcoming data scarcity and advance towards a structured web of datasets and simulators to support the development of ML for power systems.

42 ENGINEERING↗
DOE OSTI2022

Building a More Diverse, Equitable, and Inclusive Energy Efficiency Workforce

The U.S. Department of Energy (DOE) Building Technologies Office (BTO) envisions a future where the U.S. building industry leads globally in delivering quality efficiency products and services to American consumers and businesses. One barrier to achieving this vision is the difficulty reported by energy efficiency employers in finding qualified candidates (NASEO and EFI 2020). However, this barrier also presents an opportunity for the industry to expand in a way that invites and supports a more diverse, equitable, and inclusive workforce. The scope of energy, climate, and technology challenges the country faces will require the talents of all Americans. The purpose of this study therefore is to identify key factors to creating successful workforce development and inclusion programs that will support the future of the building industry. These factors can be applied to individual companies, training programs, or broader industry efforts.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2020

How Low Can You Go? Using Synthetic 3D Imagery to Drastically Reduce Real-World Training Data for Object Detection

Deep convolutional neural networks (DCNNs) currently provide state-of-the-art performance on image classification and object detection tasks, and there are many global security mission areas where such models could be extremely useful. Crucially, the success of these models is driven in large part by the widespread availability of high-quality open source data sets such as Image Net, Common Objects in Context (COCO), and KITTI, which contain millions of images with thousands of unique labels. However, global security relevant objects-of-interest can be difficult to obtain: relevant events are low frequency and high consequence; the content of relevant images is sensitive; and adversaries and proliferators seek to obscure their activities. For these cases where exemplar data is hard to come-by, even fine-tuning an existing model with available data can be effectively impossible. Recent work demonstrated that models can be trained using a combination of real-world and synthetic images generated from 3D representations; that such models can exceed the performance of models trained using real-world data alone; and that the generated images need not be perfectly realistic (Tremblay, et al., 2018). However, this approach still required hundreds to thousands of real-world images for training and fine tuning, which for sparse, global security-relevant datasets can be an unrealistic hurdle. In this research, we validate the performance and behavior of DCNN models as we drive the number of real-world images used for training object detection tasks down to a minimal set. We perform multiple experiments to identify the best approach to train DCNNs from an extremely small set of real-world images. In doing so, we: Develop state-of-the-art, parameterized 3D models based on real-world images and sample from their parameters to increase the variance in synthetic image training data; Use machine learning explainability techniques to highlight and correct through targeted training the biases that result from training using completely synthetic images; and Validate our results by comparing the performance of the models trained on synthetic data to one another, and to a control model created by fine-tuning an existing ImageNet-trained model with a limited number (hundreds) of real-world images.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Polarized and unpolarized gluon PDFs: Generative machine learning applications for lattice QCD matrix elements at short distance and large momentum

Lattice quantum chromodynamics (QCD) calculations share a defining challenge by requiring a small finite range of spatial separation z between quark/gluon bilinears for controllable power corrections in the perturbative QCD factorization, and a large hadron boost p z for a successful determination of collinear parton distribution functions (PDFs). However, these two requirements make the determination of PDFs from lattice data very challenging. We present the application of generative machine learning algorithms to estimate the polarized and unpolarized gluon correlation functions utilizing short-distance data and extending the correlation up to z p z ≲ 14 , surpassing the current capabilities of lattice QCD calculations. We train physics-informed machine learning algorithms to learn from the short-distance correlation at z ≲ 0.36 fm and take the limit, p z → ∞ , thereby minimizing possible contamination from the higher-twist effects for a successful reconstruction of the polarized gluon PDF. We also expose the bias and problems with underestimating uncertainties associated with the use of model-dependent and overly constrained functional forms, such as x α ( 1 − x ) β and its variants to extract PDFs from the lattice data. We propose the use of generative machine learning algorithms to mitigate these issues and present our determination of the polarized and unpolarized gluon PDFs in the nucleon. Published by the American Physical Society 2025

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

High Efficiency InGaN LEDs Emitting in Green, Amber and Beyond (Final Report)

This project focuses on developing high efficiency light emitting diodes (LEDs) with green, amber, and longer wavelengths using two compatible material systems based on III-nitride and II-IV-nitride semiconductor heterostructures. The proposed approach uses low-indium-content InGaN and ZnGeN 2 or ZnSnN 2 heterostructure quantum wells (QWs) as the active regions for high efficiency green and amber LEDs, and therefore to advance high efficiency color-mixed white LED technology. This approach takes advantage of the increased flexibility, afforded by the incorporation of the II-IV-nitrides into the III-nitride layers, to tune the QW active region to lower the emission wavelengths and to increase the electron-hole wave function overlap. This strategy paves a new way to extend the InGaN QW LED emission wavelength without using high-indium-content InGaN. An industrially preferable metalorganic chemical vapor deposition (MOCVD) method was used to develop the growth of ZnGeN 2 and ZnSnN 2 that are compatible with the InGaN growth. The unique dual chamber MOCVD instrument at OSU was used in this project. This program created an opportunity for the two institutions (The Ohio State University and Case Western Reserve University) with complementary expertise to perform this work. Two graduate students (Md Rezaul Karim (OSU) and Benthara Hewage Dinushi Jayatunga (CWRU)) gained experience in MOCVD growth and various characterization methods including scanning electron microscopy, x-ray diffraction spectroscopy, x-ray photoemission spectroscopy, Hall measurements, cathodoluminescence and photoluminescence. Both students successfully defended their PhD dissertations in April/May 2021. Both Karim and Jayatunga have joined Intel as engineers in June/July 2021. Another two graduate students Kaitian Zhang (OSU) and Chenxi Hu (CWRU) joined the project team since spring 2021 to continue on this project. They were trained on MOCVD epitaxy of InGaN QWs and materials characterization including SEM, AFM, PL, CL, XRD, Hall measurements and electroluminescence. A third student, Vijay Gopal Thirupakuzi Vangipuram (OSU), who joined the group in fall 2021 to perform LED fabrication and characterization. The key research accomplishments of the project include: (1) Successful development of MOCVD epitaxy of single crystalline ZnGeN 2 and ZnSnN 2 . MOCVD growth parameters that determine the stoichiometry of ZnGeN 2 and ZnSnN 2 films were identified. (2) Experimentally determined the band offsets between GaN/ZnGeN 2 and GaN/ZnGaSnN 2 hetero-interfaces, which agree very well with the theoretical predictions. (3) Demonstrated the incorporation of thin ZnGeN 2 layer in InGaN QWs. Proof-of-concept of the InGaN-ZnGeN 2 QW design demonstrated as much as 80 nm of red shift without increasing In composition or QW thickness. (4) InGaN/ZnGeN 2 QWs emitting at 480 nm achieved IQE of 47%. Resulted from this project, 7 journal articles were published, and 9 conference papers were presented.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Self-Supervised and Interpretable Anomaly Detection Using Network Transformers

Machine learning and deep neural networks (DNNs) have been proposed as a tool to identify anomalies in computer network communications. However, due the obfuscated nature of off-the-shelf machine learning models, their output often does not provide enough information to isolate the source of the anomaly to take corrective measures. In this article, we introduce the network transformer (NeT), a DNN model for anomaly detection that incorporates the graph structure of the communication network in order to improve interpretability. Further, the presented approach has the following advantages: first, enhanced interpretability by incorporating the graph structure of computer networks; second, provides a hierarchical set of features that enables analysis at different levels of granularity; second, self-supervised training that does not require labeled data. The NeT model was evaluated on a set of anomalous scenarios executed in a real industrial control system. The presented approach successfully identified the anomalies, the devices affected, and the specific connections causing the anomalies, providing a data-driven hierarchical approach to analyze the behavior of a cyber network.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

CRCNS21 Computational Models of Multisensory Integration by Upper Limb in Humanoids and Amputees

This international collaborative research project between Johns Hopkins University (JHU) and the Technical University of Munich (TUM) investigated how the human brain processes and integrates multiple types of sensory information, such as touch and force, with the goal of improving prosthetic limbs for amputees and advancing sensory capabilities in humanoid robots. The research advanced our understanding of how the brain responds to sensory feedback in upper-limb amputees. Through experiments in which amputees received electrical stimulation while performing phantom hand movements, we demonstrated that sensory feedback activates the cortical sensorimotor and multisensory regions, and that these regions communicate dynamically during stimulation. Experiments with intact-limb participants explored the integration of visual, haptic, and force feedback, as well as in virtual reality motor training, further showing how the brain processes multimodal sensory information. In addition, this research inspired work on examining the reliability of where amputees perceive sensations over time, which contributed to a successful doctoral fellowship for continued investigation. Our collaborators at TUM improved multimodal sensor technology combining tactile and thermal feedback for humanoid robots, demonstrating the feasibility of integrating multiple sensor types into a unified system for detecting and responding to environmental stimuli. The experimental methods and analysis techniques developed across both teams, including functional network analysis and multimodal sensor integration, provide a foundation for future research in prosthetics and robotics. This research benefits the public by generating knowledge about how amputees process restored sensory information. Advances in humanoid sensing contribute to safer human-robot interaction. The project also fostered international collaboration and cross-disciplinary training: one TUM doctoral student spent a summer at JHU working on multimodal sensor integration, while two JHU students traveled to TUM to host workshops on neuromorphic sensory encoding and sensory integration.

42 ENGINEERING↗