SEARCH · Engineering Papers

Results for “Generative Neural Networks”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 793 records · Page 44

All sources NASA NTRS DOE OSTI

DOE OSTI2021

Explicit physics-informed neural networks for nonlinear closure: The case of transport in tissues

In upscaling methods, closures for nonlinear problems present a well-known challenge. While a number of theoretical methods have been proposed for handling such closures, nonlinearities still remain a significant obstacle for many problems. In this work, we use a combination of formal upscaling and data-driven machine learning for explicitly closing a nonlinear transport and reaction process in multiscale tissues. The classical effectiveness factor model is used to formulate the macroscale reaction kinetics. We train a multilayer perceptron network using training data generated by direct numerical simulations over microscale examples. Once trained, the network is used in an algorithm for numerically solving the upscaled (coarse-grained) differential equation describing mass transport and reaction in two example tissues. The network is described as being explicit in the sense that the network is trained using macroscale concentrations and gradients of concentration as components of the feature space rather than incorporating them as part of a constraint in the optimization process. Network training and solutions to the macroscale transport equations were computed for two different tissues. The two tissue types (brain and liver) exhibit markedly different geometrical complexity and spatial scale (cell size and sample size). The upscaled solutions for the average concentration are compared with numerical solutions derived from the microscale concentration fields by a posteriori averaging. There are three outcomes of this work of particular note. 1) Our overall approach results in an upscaled nonlinear PDE. The PDE is closed using a neural network, and our approach results in the definition of the classical effectiveness factor for effecting closure. 2) We identify particular source terms for the closure problem that are important for representing the structure of the closure. These source terms involve macroscale concentrations and their gradients. We adopt these source terms to use as explicit features in the learning algorithm. We find the trained networks that include the macroscale source terms generate models that are able to predict the correction factor with increased fidelity over those that do not. 3) We find that the trained network exhibits good generalizability, and it is able to predict the effectiveness factor with high fidelity for realistically-structured tissues despite the significantly different scale and geometrical complexity of the two example tissue types. This latter result emphasizes our purposeful connection between conventional averaging methods with the use of machine learning for closure; this contrasts with some machine learning methods for upscaling where the exact form of the macroscale equation remains unknown.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1993

Nonlinear rescaling of control values simplifies fuzzy control

Traditional control theory is well-developed mainly for linear control situations. In non-linear cases there is no general method of generating a good control, so we have to rely on the ability of the experts (operators) to control them. If we want to automate their control, we must acquire their knowledge and translate it into a precise control strategy. The experts' knowledge is usually represented in non-numeric terms, namely, in terms of uncertain statements of the type 'if the obstacle is straight ahead, the distance to it is small, and the velocity of the car is medium, press the brakes hard'. Fuzzy control is a methodology that translates such statements into precise formulas for control. The necessary first step of this strategy consists of assigning membership functions to all the terms that the expert uses in his rules (in our sample phrase these words are 'small', 'medium', and 'hard'). The appropriate choice of a membership function can drastically improve the quality of a fuzzy control. In the simplest cases, we can take the functions whose domains have equally spaced endpoints. Because of that, many software packages for fuzzy control are based on this choice of membership functions. This choice is not very efficient in more complicated cases. Therefore, methods have been developed that use neural networks or generic algorithms to 'tune' membership functions. But this tuning takes lots of time (for example, several thousands iterations are typical for neural networks). In some cases there are evident physical reasons why equally space domains do not work: e.g., if the control variable u is always positive (i.e., if we control temperature in a reactor), then negative values (that are generated by equal spacing) simply make no sense. In this case it sounds reasonable to choose another scale u' = f(u) to represent u, so that equal spacing will work fine for u'. In the present paper we formulate the problem of finding the best rescaling function, solve this problem, and show (on a real-life example) that after an optimal rescaling, the un-tuned fuzzy control can be as good as the best state-of-art traditional non-linear controls.

Vanlangingham, H.↗

DOE OSTI2025

Structural differences between human and mouse neurons and their implementation in generative AIs

Mouse and human brains have different functions that depend on their neuronal networks. We analyzed nanometer-scale three-dimensional structures of brain tissues of the mouse medial prefrontal cortex and compared them with structures of the human anterior cingulate cortex. The obtained results indicated that mouse neuronal somata are smaller and neurites are thinner than those of human neurons. We implemented these characteristics of mouse neurons in convolutional layers of a generative adversarial network (GAN) and a denoising diffusion implicit model (DDIM), which were then subjected to image generation tasks using photo datasets of cat faces, cheese, human faces, birds, and automobiles. The mouse-mimetic GAN outperformed a standard GAN in the image generation task using the cat faces and cheese photo datasets, but underperformed for human faces and birds. The mouse-mimetic DDIM gave similar results, suggesting that the nature of the datasets affected the results. Analyses of the five datasets indicated differences in their image entropy, which should influence the number of parameters required for image generation. The preferences of the mouse-mimetic AIs coincided with the impressions commonly associated with mice. The relationship between the neuronal network and brain function should be investigated by implementing other biological findings in artificial neural networks.

generative AI↗

DOE OSTI2021

Machine-learning based prediction of injection rate and solenoid voltage characteristics in GDI injectors

We report that current state-of-the-art gasoline direct-injection (GDI) engines use multiple injections as one of the key technologies to improve exhaust emissions and fuel efficiency. For this technology to be successful, secured adequate control of fuel quantity for each injection is mandatory. However, nonlinearity and variations in the injection quantity can deteriorate the accuracy of fuel control, especially with small fuel injections. Therefore, it is necessary to understand the complex injection behavior and to develop a predictive model to be utilized in the development process. This study presents a methodology for rate of injection (ROI) and solenoid voltage modeling using artificial neural networks (ANNs) constructed from a set of Zeuch-style hydraulic experimental measurements conducted over a wide range of conditions. A quantitative comparison between the ANN model and the experimental data shows that the model is capable of predicting not only general features of the ROI trend, but also transient and non-linear behaviors at particular conditions. In addition, the end of injection (EOI) could be detected precisely with a virtually generated solenoid voltage signal and the signal processing method, which applies to an actual engine control unit. A correlation between the detected EOI timings calculated from the modeled signal and the measurement results showed a high coefficient of determination.

42 ENGINEERING↗

DOE OSTI2023

Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware

In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications. However, activation function calculations are still being implemented with general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access memory (Ag-CBRAM) crossbar arrays with Mott rectified linear unit (ReLU) activation neurons for scalable, energy and area-efficient hardware (HW) implementation of deep neural networks. We develop Ag-CBRAM devices that can achieve a high ON/OFF ratio and multi-level programmability. Compact and energy-efficient Mott ReLU neuron devices implementing ReLU activation function are directly connected to the columns of Ag-CBRAM crossbars to compute the output from the weighted sum current. We implement convolution filters and activations for VGG-16 using our integrated HW and demonstrate the successful generation of feature maps for CIFAR-10 images in HW. Our approach paves a new way toward building a highly compact and energy-efficient eNVMs-based in-memory computing system.

Mott insulators↗

DOE OSTI2023

Denoising diffusion algorithm for inverse design of microstructures with fine-tuned nonlinear material properties

Here we introduce a denoising diffusion algorithm to discover microstructures with nonlinear fine-tuned properties. Denoising diffusion probabilistic models are generative models that use diffusion-based dynamics to gradually denoise images and generate realistic synthetic samples. By learning the reverse of a Markov diffusion process, we design an artificial intelligence to efficiently manipulate the topology of microstructures to generate a massive number of prototypes that exhibit constitutive responses sufficiently close to designated nonlinear constitutive behaviors. To identify the subset of microcstructures with sufficiently precise fine-tuned properties, a convolutional neural network surrogate is trained to replace high-fidelity finite element simulations to filter out prototypes outside the admissible range. Results of this study indicate that the denoising diffusion process is capable of creating microstructures of fine-tuned nonlinear material properties within the latent space of the training data. More importantly, this denoising diffusion algorithm can be easily extended to incorporate additional topological and geometric modifications by introducing high-dimensional structures embedded in the latent space. Numerical experiments are conducted on the open-source mechanical MNIST data set (Lejeune, 2020). Consequently, this algorithm is not only capable of performing inverse design of nonlinear effective media, but also learns the nonlinear structure–property map to quantitatively understand the multiscale interplay among the geometry, topology, and their effective macroscopic properties.

42 ENGINEERING↗

DOE OSTI2023

Feature-adjacent multi-fidelity physics-informed machine learning for partial differential equations

Physics-informed neural networks have emerged as an alternative method for solving partial differential equations. However, for complex problems, the training of such networks can still require high-fidelity data which can be expensive to generate. To reduce or even eliminate the dependency on high-fidelity data, we propose a novel multi-fidelity architecture which is based on a feature space shared by the low- and high-fidelity solutions. In the feature space, the projections of the low-fidelity and high-fidelity solutions are adjacent by constraining their relative distance. The feature space is represented with an encoder and its mapping to the original solution space is effected through a decoder. In conclusion, the proposed multi-fidelity approach is validated on forward and inverse problems for steady and unsteady problems described by partial differential equations.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Integrated analysis of X-ray diffraction patterns and pair distribution functions for machine-learned phase identification

Abstract To bolster the accuracy of existing methods for automated phase identification from X-ray diffraction (XRD) patterns, we introduce a machine learning approach that uses a dual representation whereby XRD patterns are augmented with simulated pair distribution functions (PDFs). A convolutional neural network is trained directly on XRD patterns calculated using physics-informed data augmentation, which accounts for experimental artifacts such as lattice strain and crystallographic texture. A second network is trained on PDFs generated via Fourier transform of the augmented XRD patterns. At inference, these networks classify unknown samples by aggregating their predictions in a confidence-weighted sum. We show that such an integrated approach to phase identification provides enhanced accuracy by leveraging the benefits of each model’s input representation. Whereas networks trained on XRD patterns provide a reciprocal space representation and can effectively distinguish large diffraction peaks in multi-phase samples, networks trained on PDFs provide a real space representation and perform better when peaks with low intensity become important. These findings underscore the importance of using diverse input representations for machine learning models in materials science and point to new avenues for automating multi-modal characterization.

36 MATERIALS SCIENCE↗

DOE OSTI2023

Characterizing Quantum Classifier Utility in Natural Language Processing Workflows

Quantum Natural Language Processing (QNLP) develops natural language processing (NLP) models for deployment on quantum computers. We explore feature and data prototype selection techniques to address challenges posed by encoding high dimensional features. Our study builds quantum circuit classifiers that includes classical feature pre-processing, quantum embedding and quantum model training. The quantum models are built on 4 or 6 qubits and the quantum neural network (QNN) uses the established bricklayer design. We compare the dependence of model performance (in terms of accuracy and F1 scores) on feature length, embedding gates and parameterized unitary design. We compare the performance of quantum machine learning models to classical convolution neural network model (CNN) on binary and multi-class classification tasks using two datasets of synthetic features and labels. The first is the ECP-CANDLE P3B3 dataset a corpus of synthetically generated cancer pathology reports. The second dataset is extracted from well-known benchmark dataset (MADELON) - features are generated with a combination of informative, repeated and uninformative features. Both datasets are used for binary classification and multi-class classification with 3 classes. We observe robust, accurate performance from all models on the binary classification tasks, but multiclass classification is a challenge for the quantum models-there is a notable decrease in accuracy when using 3 classes. Overall the performance is comparable in terms of recall and accuracy between QNNs and CNNs, even with large datasets. These results provide a point of comparison between quantum and classical models on real-world datasets.

Hamilton, Kathleen↗

NASA NTRS1996

Cloud Classification in Polar and Desert Regions and Smoke Classification from Biomass Burning Using a Hierarchical Neural Network

This research focuses on a new neural network scene classification technique. The task is to identify scene elements in Advanced Very High Resolution Radiometry (AVHRR) data from three scene types: polar, desert and smoke from biomass burning in South America (smoke). The ultimate goal of this research is to design and implement a computer system which will identify the clouds present on a whole-Earth satellite view as a means of tracking global climate changes. Previous research has reported results for rule-based systems (Tovinkere et at 1992, 1993) for standard back propagation (Watters et at. 1993) and for a hierarchical approach (Corwin et al 1994) for polar data. This research uses a hierarchical neural network with don't care conditions and applies this technique to complex scenes. A hierarchical neural network consists of a switching network and a collection of leaf networks. The idea of the hierarchical neural network is that it is a simpler task to classify a certain pattern from a subset of patterns than it is to classify a pattern from the entire set. Therefore, the first task is to cluster the classes into groups. The switching, or decision network, performs an initial classification by selecting a leaf network. The leaf networks contain a reduced set of similar classes, and it is in the various leaf networks that the actual classification takes place. The grouping of classes in the various leaf networks is determined by applying an iterative clustering algorithm. Several clustering algorithms were investigated, but due to the size of the data sets, the exhaustive search algorithms were eliminated. A heuristic approach using a confusion matrix from a lightly trained neural network provided the basis for the clustering algorithm. Once the clusters have been identified, the hierarchical network can be trained. The approach of using don't care nodes results from the difficulty in generating extremely complex surfaces in order to separate one class from all of the others. This approach finds pairwise separating surfaces and forms the more complex separating surface from combinations of simpler surfaces. This technique both reduces training time and improves accuracy over the previously reported results. Accuracies of 97.47%, 95.70%, and 99.05% were achieved for the polar, desert and smoke data sets.

Alexander, June↗

NASA NTRSDate not supplied

Hailstorm Analyses and Detection Derived from Current and Historical Satellite Data and Convective Environmental Parameters

We seek to demonstrate the extent to which hailstorms can be detected using a combination of geostationary (GEO) visible and infrared metrics of storm intensity and convective environmental parameters from reanalysis. Hailstorm identifications from low-Earth-orbiting (LEO) passive microwave sensors and maximum expected size of hail (MESH) from ground-based radar serve as a proxy for hail events. Data are analyzed for two warm seasons over the contiguous United States (CONUS). A neural network (NN) is trained to predict hailstorm detection dependent on optimal multi-variate weighting of observed and modeled input. The NN results are then applied to a 15-year Meteosat Second Generation climatology over South Africa to depict where hailstorms are most likely to occur. We also explore the impact of GEO imager resolution on our ability to discriminate hailstorms by matching GOES-13 (4 km) and GOES-16 (2 km) storm intensity metrics against hail characteristics using LEO, MESH, and spotter reports over CONUS during 2017, when both satellites were simultaneously imaging. Such analysis allows us to assess the feasibility of assembling a severe storm climate data record back to GOES-8, in the mid-1990’s. Finer spatial resolution of GOES-16 better resolves the updraft characteristics and intensities that are inherently linked to hail formation; however, GOES-13 can be normalized to achieve comparable detection capability. When intelligently combined with model-derived convective environmental parameters, GEO-derived storm intensity metrics enable high-spatial resolution hail risk assessment at hourly intervals throughout the diurnal cycle anywhere around the world and an improved understanding of hailstorms in the climate system.

Kyle F Itterly↗

NASA NTRS2026

Proactive Wildfire Management: A Remote Sensing and Multimodal CNN-MLP Architecture for Ignition Risk Forecasting

As the frequency and intensity of wildfires increase, with fire seasons now starting earlier and ending later than they have over the past decades, current monitoring systems, such as lookout towers and satellites, are hindered by cloud cover, low-resolution imagery, and static data gaps that fail to track vegetation moisture levels fast enough to catch rapid pre-ignition changes. This report proposes a Machine Learning-enabled Wildfire Ignition Prediction framework that combines satellite monitoring with dynamic and high-resolution remote sensing from Unmanned Aerial Vehicle (UAV) swarms. The method would use multispectral and thermal data from the Landsat program to create a baseline for vegetation health, calculating a two-band Enhanced Vegetation Index (EVI2) and the moisture content of the vegetation. These inputs will later be fused with microscale UAV weather data, including thermal hotspots found through thick canopies, hyperspectral chemical signatures of pre-visual combustion, and local weather streams. The multispectral satellite, multispectral Light Detection and Ranging (LiDAR), and thermal data would then be processed through a Convolutional Neural Network (CNN), alongside a Multilayer Perceptron (MLP) for the micro-weather telemetry. The outputs of these networks would be fused into a single feature representation and passed through a final prediction network to generate real-time ignition risk scores and hotspot alerts. Model performance would be assessed using standard classification metrics, including a Receiver Operating Characteristic - Area Under the Curve (ROC AUC) and F1 score. This system would allow first responders to identify high-risk zones and intervene before ignition occurs, improving emergency response time compared to current approaches.

machine learning↗

DOE OSTI2022

Hyperspectral remote sensing-based plant community map for region around NGEE-Arctic intensive research watersheds at Seward Peninsula, Alaska, 2017-2019

Using airborne hyperspectral remote sensing data from NASA Airborne Visible-Infrared Imaging Spectrometer- Next Generation (AVIRIS-NG) platforms in a region near NGEE-Arctic intensive watersheds at Seward peninsula of Alaska, high resolution (5m) maps of plant community distribution were developed and included in this data collected. AVIRIS-NG data collected over 2017-2019 period were used to develop deep neural networks, trained using vegetation plot observations collected at NGEE-Arctic watersheds at Kougarok, Council and Teller. A hierarchical vegetation classification scheme consisting of six classes at Level I, and 16 classes at Level II contained in two .txt files were used to developed the plant community maps for the region. Two geospatial raster data files (.tif) at both thematic levels are shared in this data collection. Data files in this collection use Alaska Albers Equal Area projection. Readme files available in three formats (*.html, *.md, *.pdf) and one *.png visualization map.The Next-Generation Ecosystem Experiments: Arctic (NGEE Arctic), was a research effort to reduce uncertainty in Earth System Models by developing a predictive understanding of carbon-rich Arctic ecosystems and feedbacks to climate. NGEE Arctic was supported by the Department of Energy's Office of Biological and Environmental Research.The NGEE Arctic project had two field research sites: 1) located within the Arctic polygonal tundra coastal region on the Barrow Environmental Observatory (BEO) and the North Slope near Utqiagvik (Barrow), Alaska and 2) multiple areas on the discontinuous permafrost region of the Seward Peninsula north of Nome, Alaska.Through observations, experiments, and synthesis with existing datasets, NGEE Arctic provided an enhanced knowledge base for multi-scale modeling and contributed to improved process representation at global pan-Arctic scales within the Department of Energy's Earth system Model (the Energy Exascale Earth System Model, or E3SM), and specifically within the E3SM Land Model component (ELM).

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

An Investigation of LES Wall Modeling for Rayleigh–Bénard Convection via Interpretable and Physics-Aware Feedforward Neural Networks with DNS

Abstract The traditional approach of using the Monin–Obukhov similarity theory (MOST) to model near-surface processes in large-eddy simulations (LESs) can lead to significant errors in natural convection. In this study, we propose an alternative approach based on feedforward neural networks (FNNs) trained on output from direct numerical simulation (DNS). To evaluate the performance, we conduct both a priori and a posteriori tests. In the a priori (offline) tests, we compare the statistics of the surface shear stress and heat flux, computed from filtered DNS input variables, to the stress and flux obtained from the filtered DNS. Additionally, we investigate the importance of various input features using the Shapley additive explanations value and the conditional average of the filter grid cells. In the a posteriori (online) tests, we implement the trained models in the System for Atmospheric Modeling (SAM) LES and compare the LES-generated surface shear stress and heat flux with those in the DNS. Our findings reveal that vertical velocity, a traditionally overlooked flow quantity, is one of the most important input features for determining the wall fluxes. Increasing the number of input features improves the a priori test results but does not always improve the model performance in the a posteriori tests because of the differences in input variables between the LES and DNS. Last, we show that physics-aware FNN models trained with logarithmic and scaled parameters can well extrapolate to more intense convection scenarios than in the training dataset, whereas those trained with primitive flow quantities cannot. Significance Statement The traditional near-surface turbulence model, based on a shear-dominated boundary layer flow, does not represent near-surface turbulence in natural convection. Using a feedforward neural network (FNN), we can construct a more accurate model that better represents the near-surface turbulence in various flows and reveals previously overlooked controlling factors and process interactions. Our study shows that the FNN-generated models outperform the traditional model and highlight the importance of the near-surface vertical velocity. Furthermore, the physics-aware FNN models exhibit the potential to extrapolate to convective flows of various intensities beyond the range of the training dataset, suggesting their broader applicability for more accurate modeling of near-surface turbulence.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS2005

Automated Wildfire Detection Through Artificial Neural Networks

Wildfires have a profound impact upon the biosphere and our society in general. They cause loss of life, destruction of personal property and natural resources and alter the chemistry of the atmosphere. In response to the concern over the consequences of wildland fire and to support the fire management community, the National Oceanic and Atmospheric Administration (NOAA), National Environmental Satellite, Data and Information Service (NESDIS) located in Camp Springs, Maryland gradually developed an operational system to routinely monitor wildland fire by satellite observations. The Hazard Mapping System, as it is known today, allows a team of trained fire analysts to examine and integrate, on a daily basis, remote sensing data from Geostationary Operational Environmental Satellite (GOES), Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS) satellite sensors and generate a 24 hour fire product for the conterminous United States. Although assisted by automated fire detection algorithms, N O M has not been able to eliminate the human element from their fire detection procedures. As a consequence, the manually intensive effort has prevented NOAA from transitioning to a global fire product as urged particularly by climate modelers. NASA at Goddard Space Flight Center in Greenbelt, Maryland is helping N O M more fully automate the Hazard Mapping System by training neural networks to mimic the decision-making process of the frre analyst team as well as the automated algorithms.

Miller, Jerry↗

DOE OSTI2025

Explainable Synthesizability Prediction of Inorganic Crystal Polymorphs Using Large Language Models

Abstract We evaluate the ability of machine learning to predict whether a hypothetical crystal structure can be synthesized and explain those predictions to scientists. Fine‐tuned large language models (LLMs) trained on a human‐readable text description of the target crystal structure perform comparably to previous bespoke convolutional graph neural network methods, but better prediction quality can be achieved by training a positive‐unlabeled learning model on a text‐embedding representation of the structure. An LLM‐based workflow can then be used to generate human‐readable explanations for the types of factors governing synthesizability, extract the underlying physical rules, and assess the veracity of those rules. These explanations can guide chemists in modifying or optimizing non‐synthesizable hypothetical structures to make them more feasible for materials design.

Kim, Seongmin [Department of Chemical and Biologic↗

DOE OSTI2025

Explainable Synthesizability Prediction of Inorganic Crystal Polymorphs Using Large Language Models

Kim, Seongmin [Department of Chemical and Biologic↗

DOE OSTI2026

PowerModelsGAT-AI: Physics-Informed Graph Attention for Multi-System Power Flow With Continual Learning

Solving the alternating current power flow equations in real time is essential for secure grid operation, yet classical Newton–Raphson solvers can be slow under stressed conditions. Existing graph neural networks for power flow are typically trained on a single system and often degrade on different systems. We present PowerModelsGAT-AI, a physics-informed graph attention network that predicts bus voltages and generator injections. The model uses bus-type-aware masking to handle different bus types and balances multiple loss terms, including a power-mismatch penalty, using learned weights. We evaluate the model on 14 benchmark systems (4 to 6,470 buses) and train a unified model on 13 of these under contingency conditions with up to two branch outages, achieving an average normalized mean absolute error of 0.89% for voltage magnitudes and R 2 >0.99 for voltage angles. We also show continual learning: when adapting a base model to a new 1,354-bus system, standard fine-tuning causes severe forgetting with error increases exceeding 1000% on base systems, while our experience replay and elastic weight consolidation strategy keeps error increases below 2% and in some cases improves base-system performance. Interpretability analysis shows that learned attention weights correlate with physical branch parameters (susceptance: r=0.38 ; thermal limits: r=0.22 ), and feature importance analysis supports that the model captures established power flow relationships.

24 POWER TRANSMISSION AND DISTRIBUTION↗