Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Deep Neural Network”

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 1,117 records · Page 62
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Photometric identification of compact galaxies, stars, and quasars using multiple neural networks

We present MargNet, a deep learning-based classifier for identifying stars, quasars, and compact galaxies using photometric parameters and images from the Sloan Digital Sky Survey Data Release 16 catalogue. MargNet consists of a combination of convolutional neural network and artificial neural network architectures. Using a carefully curated data set consisting of 240 000 compact objects and an additional 150 000 faint objects, the machine learns classification directly from the data, minimizing the need for human intervention. MargNet is the first classifier focusing exclusively on compact galaxies and performs better than other methods to classify compact galaxies from stars and quasars, even at fainter magnitudes. This model and feature engineering in such deep learning architectures will provide greater success in identifying objects in the ongoing and upcoming surveys, such as Dark Energy Survey and images from the Vera C. Rubin Observatory.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Neural-network model for force prediction in multi-principal-element alloys

Atomistic simulations can provide useful insights into the physical properties of multi-principal-element alloys. However, classical potentials mostly fail to capture key quantum (electronic-structure) effects. Here we present a deep 3D convolutional neural network (3D CNN) based framework combined with a voxelization technique to design interatomic potentials for chemically complex alloys. We highlight the performance of the 3D CNN model and its efficacy in computing potentials using the medium-entropy alloy TaNbMo. In order to provide insights into the effect of voxel resolution, we implemented two approaches based on the inner and outer bounding boxes. An efficient 3D CNN model, which is as accurate as the density-functional theory (DFT) approach, for calculating potentials will provide a promising schema for accurate atomistic simulations of structure and dynamics of general multi-principle element alloys.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Abisko: Deep codesign of an architecture for spiking neural networks using novel neuromorphic materials

The Abisko project aims to develop an energy-efficient spiking neural network (SNN) computing architecture and software system capable of autonomous learning and operation. The SNN architecture explores novel neuromorphic devices that are based on resistive-switching materials, such as memristors and electrochemical RAM. Equally important, Abisko uses a deep codesign approach to pursue this goal by engaging experts from across the entire range of disciplines: materials, devices and circuits, architectures and integration, software, and algorithms. Here, the key objectives of our Abisko project are threefold. First, we are designing an energy-optimized high-performance neuromorphic accelerator based on SNNs. This architecture is being designed as a chiplet that can be deployed in contemporary computer architectures and we are investigating novel neuromorphic materials to improve its design. Second, we are concurrently developing a productive software stack for the neuromorphic accelerator that will also be portable to other architectures, such as field-programmable gate arrays and GPUs. Third, we are creating a new deep codesign methodology and framework for developing clear interfaces, requirements, and metrics between each level of abstraction to enable the system design to be explored and implemented interchangeably with execution, measurement, a model, or simulation. As a motivating application for this codesign effort, we target the use of SNNs for an analog event detector for a high-energy physics sensor.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Deep learning models for interpretation of point of care ultrasound in military working dogs

Introduction: Military working dogs (MWDs) are essential for military operations in a wide range of missions. With this pivotal role, MWDs can become casualties requiring specialized veterinary care that may not always be available far forward on the battlefield. Some injuries such as pneumothorax, hemothorax, or abdominal hemorrhage can be diagnosed using point of care ultrasound (POCUS) such as the Global FAST® exam. This presents a unique opportunity for artificial intelligence (AI) to aid in the interpretation of ultrasound images. In this article, deep learning classification neural networks were developed for POCUS assessment in MWDs. Methods: Images were collected in five MWDs under general anesthesia or deep sedation for all scan points in the Global FAST® exam. For representative injuries, a cadaver model was used from which positive and negative injury images were captured. A total of 327 ultrasound clips were captured and split across scan points for training three different AI network architectures: MobileNetV2, DarkNet-19, and ShrapML. Gradient class activation mapping (GradCAM) overlays were generated for representative images to better explain AI predictions. Results: Performance of AI models reached over 82% accuracy for all scan points. The model with the highest performance was trained with the MobileNetV2 network for the cystocolic scan point achieving 99.8% accuracy. Across all trained networks the diaphragmatic hepatorenal scan point had the best overall performance. However, GradCAM overlays showed that the models with highest accuracy, like MobileNetV2, were not always identifying relevant features. Conversely, the GradCAM heatmaps for ShrapML show general agreement with regions most indicative of fluid accumulation. Discussion: Overall, the AI models developed can automate POCUS predictions in MWDs. Preliminarily, ShrapML had the strongest performance and prediction rate paired with accurately tracking fluid accumulation sites, making it the most suitable option for eventual real-time deployment with ultrasound systems. Further integration of this technology with imaging technologies will expand use of POCUS-based triage of MWDs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Learning nuclear cross sections across the chart of nuclides with graph neural networks

We explore the use of deep learning techniques to learn how nuclear cross sections change as we add or remove protons and neutrons. As a proof of principle, we focus on the neutron-induced reactions in the fast energy regime. Our approach follows a two-stage learning framework. First, we apply representation learning to encode cross section data into a latent space using either variational autoencoders (VAEs) or implicit neural representations (INRs). Then, we train graph neural networks (GNNs) on the resulting embeddings to predict missing values across the nuclear chart by leveraging the topological structure of neighboring isotopes. We demonstrate accurate cross section predictions within a 9 × 9 block of missing nuclei. We also find that the optimal GNN training strategy depends on the type of latent representation used, with VAE embeddings performing best under end-to-end optimization in the original space, while INR embeddings achieve better results when the GNN is trained only in the latent space. Furthermore, using clustering algorithms, we map groups of latent vectors into regions of the nuclear chart and show that VAEs and INRs can discover some of the neutron magic numbers. These findings suggest that deep-learning models based on the representation encoding of cross sections combined with graph neural networks hold significant potential in augmenting nuclear theory models, e.g., by providing reliable estimates of covariances of cross sections, including cross-material covariances.

Machine learning↗
DOE OSTI2025

Inference of three-dimensional hot-spot and shell morphology in inertial confinement fusion experiments using a convolutional neural network

The performance of inertial confinement fusion (ICF) implosions is sensitive to the three-dimensional (3D) morphology of the hot-spot and shell configurations. The ability to infer shell-mass uniformity and reconstruct 3D hot spots is crucial for quantifying the degradation of ignition criteria and improving symmetry in ICF implosion experiments. In this work, we present a deep-learning convolutional neural network (CNN) for reconstructing 3D hot-spot and shell structures for ICF capsules. The 3D geometry of the hot spot is reconstructed from x-ray images measured from multiple lines of sight on OMEGA. The shell configuration is inferred indirectly through machine learning using a convolutional neural network extensively trained on a dec3d simulation database. This simulation-dependent approach yields consistent agreement between reconstructed 3D shell densities and machine-learning optimized dec3d simulation results. This work demonstrates a CNN framework that successfully reconstructs 3D capsule structures from two-dimensional images in ICF implosions.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Development of a Deep Learning Model for Predicting the Drag Coefficients of Spherical and Non-Spherical Particles,

There is yet to be a well-established drag model for non-spherical particles required in a particle-laden flow that could cover a wide range of sphericities. This talk will explore the development of a general drag model for non-spherical particles by applying deep learning using available experimental data available in the literature. The integration of several raw experimental measurements from different sources and research directions allows the training of robust Artificial Intelligence and Machine Learning (ML) models. Neural networks are an ML approach inspired by the inner biological workings of the brain. This work aims to develop a Deep Neural Network (DNN) that predicts drag coefficient values with the ability to adapt appropriately to unseen data. Given the limited number of data points available and the variance found within the data collected from various sources, challenges may arise when looking to train the model. Our study tests and implements various model regularization techniques and assesses different loss and activation functions for the proposed DNN. The proposed model considers a broader range of features other than sphericity and Reynold number. These features include density ratio, solid volume fraction, lengthwise and crosswise sphericity, and more. Furthermore, we present the features that play a significant role in predicting different drag coefficients through feature importance. Within the investigated parameter ranges in this study, the following conclusions can be achieved and summarized below: • An improved drag coefficient model can be developed by considering more features such as, aspect ratio, lengthwise sphericity, crosswise sphericity, and density ratio. • DNN model can predict better results compared to traditional methods using MAE metric. • The proposed model addresses data challenges such as limited data and extreme data points through expanded feature-set and regularization. • Three major features that mostly affect the drag coefficient were identified from a feature importance analysis.

Presa-Reyes, Maria↗
DOE OSTI2024

Improving neutrino energy estimation of charged-current interaction events with recurrent neural networks in MicroBooNE

We present a deep learning-based method for estimating the neutrino energy of charged-current neutrino-argon interactions. We employ a recurrent neural network (RNN) architecture for neutrino energy estimation in the MicroBooNE experiment, utilizing liquid argon time projection chamber (LArTPC) detector technology. Traditional energy estimation approaches in LArTPCs, which largely rely on reconstructing and summing visible energies, often experience sizable biases and resolution smearing because of the complex nature of neutrino interactions and the detector response. The estimation of neutrino energy can be improved after considering the kinematics information of reconstructed final-state particles. Utilizing kinematic information of reconstructed particles, the deep learning-based approach shows improved resolution and reduced bias for the muon neutrino Monte Carlo simulation sample compared to the traditional approach. In order to address the common concern about the effectiveness of this method on experimental data, the RNN-based energy estimator is further examined and validated with dedicated data-simulation consistency tests using MicroBooNE data. We also assess its potential impact on a neutrino oscillation study after accounting for all statistical and systematic uncertainties and show that it enhances physics sensitivity. This method has good potential to improve the performance of other physics analyses. Published by the American Physical Society 2024

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2024

Speeding up and reducing memory usage for scientific machine learning via mixed precision

Scientific machine learning (SciML) has emerged as a versatile approach to address complex computational science and engineering problems. Within this field, physics-informed neural networks (PINNs) and deep operator networks (DeepONets) stand out as the leading techniques for solving partial differential equations by incorporating both physical equations and experimental data. However, training PINNs and DeepONets require significant computational resources, including long computational times and large amounts of memory. In search of computational efficiency, training neural networks using half precision (float16) rather than the conventional single (float32) or double (float64) precision has gained substantial interest, given the inherent benefits of reduced computational time and memory consumed. However, we find that float16 cannot be applied to SciML methods, because of gradient divergence at the start of training, weight updates going to zero, and the inability to converge to a local minima. To overcome these limitations, we explore mixed precision, which is an approach that combines the float16 and float32 numerical formats to reduce memory usage and increase computational speed. Our experiments showcase that mixed precision training not only substantially decreases training times and memory demands but also maintains model accuracy. Here, we also reinforce our empirical observations with a theoretical analysis. The research has broad implications for SciML in various computational applications.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Femtojoule optical nonlinearity for deep learning with incoherent illumination

Optical neural networks (ONNs) are a promising computational alternative for deep learning due to their inherent massive parallelism for linear operations. However, the development of energy-efficient and highly parallel optical nonlinearities, a critical component in ONNs, remains an outstanding challenge. Here, we introduce a nonlinear optical microdevice array (NOMA) compatible with incoherent illumination by integrating the liquid crystal cell with silicon photodiodes at the single-pixel level. We fabricate NOMA with more than half a million pixels, each functioning as an optical analog of the rectified linear unit at ultralow switching energy down to 100 femtojoules per pixel. With NOMA, we demonstrate an optical multilayer neural network. Our work holds promise for large-scale and low-power deep ONNs, computer vision, and real-time optical image processing.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

Understanding the relationship between protein sequence and function is necessary to design new and useful proteins with applications in bioenergy, medicine, and agriculture. The mapping from sequence to function is tremendously complex because it involves thousands of molecular interactions that are coupled over multiple lengths and timescales. We show that neural networks can learn the sequence–function mapping from large protein datasets. Neural networks are appealing for this task because they can learn complicated relationships from data, make few assumptions about the nature of the sequence–function relationship, and can learn general rules that apply across the length of the protein sequence. We demonstrate that learned models can be applied to design new proteins with properties that exceed natural sequences.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

ECNet is an evolutionary context-integrated deep learning framework for protein engineering

Abstract Machine learning has been increasingly used for protein engineering. However, because the general sequence contexts they capture are not specific to the protein being engineered, the accuracy of existing machine learning algorithms is rather limited. Here, we report ECNet (evolutionary context-integrated neural network), a deep-learning algorithm that exploits evolutionary contexts to predict functional fitness for protein engineering. This algorithm integrates local evolutionary context from homologous sequences that explicitly model residue-residue epistasis for the protein of interest with the global evolutionary context that encodes rich semantic and structural features from the enormous protein sequence universe. As such, it enables accurate mapping from sequence to function and provides generalization from low-order mutants to higher-order mutants. We show that ECNet predicts the sequence-function relationship more accurately as compared to existing machine learning algorithms by using ~50 deep mutational scanning and random mutagenesis datasets. Moreover, we used ECNet to guide the engineering of TEM-1 β-lactamase and identified variants with improved ampicillin resistance with high success rates.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Power System Waveform Classification Using Time-Frequency and CNN

Many modern reclosers and circuit breakers have microprocessor relays that record waveforms of system events. In some cases, utilities may record a half-a-dozen event captures for every event. This is thousands of events per year. The industry needs faster, more automated, more conclusive, and easy-to-use systems that can process massive amounts of event recordings without extensive input/support from power system engineers. To address the need for a commercially viable solution that can classify waveform data, energies were directed to develop a universal neural network (NN) structure (deep learning algorithm) that works for a wide variety of system event types. The structure that showed the most promise was one that included the use of spectrograms. The technique has shown positive results in audio engineering, particularly with respect to speech recognition. A waveform signature could be treated as a spoken word like audio waveforms for specific things such as “YES” or “UP”. No two people produce the exact same waveform when speaking each of these words, but audio processing algorithms based on spectrograms and convolutional neural networks (CNN) can still distinguish the word regardless of the speaker. No two circuits produce the exact same waveform for a given event, but the NN can be trained to classify the event type regardless of the circuit or location on the circuit. A Power System Neural Network (PSNN) has been developed to use a CNN to classify events within waveform data for power systems. The waveform is converted to an array of values by way of spectrograms and interpreted as an image. This image is passed into the CNN. The test results on independent simulated test and validation datasets show greater than 99% accuracy. While the results thus far are based on simulated data, the performance of the PSNN is very promising and should work for a wide variety of power system conditions of interest. Ultimately, much of the custom code and tools used today and much of the manual effort expended today may be automated using this PSNN.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

YOLO for Radio Frequency Signal Classification

Radio frequency signal classification plays a pivotal role in various applications, including spectrum management, wireless security, and cognitive radio. Extant signal classification methods require significant data throughput and are not multilabel. We propose a novel approach to radio frequency signal classification by leveraging the You Only Look Once (YOLO) object detection method. YOLO is a state-of-the-art deep learning model renowned for its real-time object detection capabilities in computer vision applications. We adapt YOLO for signal classification to enable the automatic and efficient identification of various signal types within a power spectral density image. Index Terms—radio-frequency analysis, object detection, neural networks, machine learning, deep learning.

42 ENGINEERING↗
NASA NTRS2024

Effective structural impact detection and localization using convolutional neural network and Bayesian information fusion with limited sensors

Due to their unpredictable nature, many impact events (e.g., overheight vehicles striking on bridges) go unnoticed or get reported many hours later. However, they can induce structural failures or hidden damage that accelerates the structure’s long-term degradation. Therefore, prompt impact detection and localization strategies are essential for early warning of impact events and rapid maintenance of structures. Most existing impact detection strategies are developed for aircraft composite panels utilizing high-rate synchronized measurement from densely deployed sensors. Limited efforts have been made for infrastructure or human habitats which generally require large-scale but low-rate measurement. In particular, due to harsh environments (e.g., deep space habitats under meteoroids), structural impact localization must be robust to limited sensors (e.g., sensor damage during impacts) and multi-source errors (e.g., measurement errors). In this study, an effective impact detection and localization strategy is proposed using a limited number of vibration measurements, especially in harsh environments (e.g. in deep space). Convolutional neural networks are trained for each sensor node and are fused using Bayesian theory to improve the accuracy of impact localization. Special considerations are paid to evaluate the effect of both measurement error and modeling error in the analysis. The proposed strategy is illustrated using 1D structure, and further validated in 3D geodesic dome structure numerically. The results demonstrate that it can detect and localize impact events accurately and robustly on structures.

Yuguang Fu↗
DOE OSTI2022

Monitoring and prediction of porosity in laser powder bed fusion using physics-informed meltpool signatures and machine learning

In this work we accomplished the monitoring and prediction of porosity in laser powder bed fusion (LPBF) additive manufacturing process. This objective was realized by extracting physics-informed meltpool signatures from an in-situ dual-wavelength imaging pyrometer, and subsequently, analyzing these signatures via computationally tractable machine learning approaches. Porosity in LPBF occurs despite extensive optimization of processing conditions due to stochastic causes. Hence, it is essential to continually monitor the process with in-situ sensors for detecting and mitigating incipient pore formation. In this work a tall cuboid-shaped part (10 mm × 10 mm × 137 mm, material ATI 718Plus) was built with controlled porosity by varying laser power and scanning speed. This test caused various types of porosity, such as lack-of-fusion and keyhole formation, with varying degrees of severity in the part. The meltpool was continuously monitored using a dual-wavelength imaging pyrometer installed in the machine. Physically intuitive process signatures, such as meltpool length, temperature distribution, and ejecta (spatter) characteristics, were extracted from the meltpool images. Subsequently, relatively simple machine learning models, e.g., K-Nearest Neighbors, were trained to predict both the severity and type of porosity as a function of these physics-informed meltpool signatures. These models resulted in a prediction accuracy exceeding 95% (statistical F1-score). The same analysis was carried out with a complex, black-box deep learning convolutional neural network which directly used the meltpool images instead of physics-informed features. The convolutional neural network produced a comparable F1-score in the range of 89–97%. Finally, these results demonstrate that using pragmatic, physics-informed meltpool signatures within a simple machine learning model is as effective for flaw prediction in LPBF as using a complex and computationally demanding black-box deep learning model.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Automated Grain Boundary (GB) Segmentation and Microstructural Analysis in 347H Stainless Steel Using Deep Learning and Multimodal Microscopy

Austenitic 347H stainless steel offers superior mechanical properties and corrosion resistance required for extreme operating conditions such as high temperature. The change in microstructure due to composition and process variations is expected to impact material properties. Identifying microstructural features such as grain boundaries thus becomes an important task in the process-microstructure-properties loop. Applying convolutional neural network (CNN)-based deep learning models is a powerful technique to detect features from material micrographs in an automated manner. In contrast to microstructural classification, supervised CNN models for segmentation tasks require pixel-wise annotation labels. However, manual labeling of the images for the segmentation task poses a major bottleneck for generating training data and labels in a reliable and reproducible way within a reasonable timeframe. Microstructural characterization especially needs to be expedited for faster material discovery by changing alloy compositions. Here, in this study, we attempt to overcome such limitations by utilizing multimodal microscopy to generate labels directly instead of manual labeling. We combine scanning electron microscopy images of 347H stainless steel as training data and electron backscatter diffraction micrographs as pixel-wise labels for grain boundary detection as a semantic segmentation task. The viability of our method is evaluated by considering a set of deep CNN architectures. We demonstrate that despite producing instrumentation drift during data collection between two modes of microscopy, this method performs comparably to similar segmentation tasks that used manual labeling. Additionally, we find that naïve pixel-wise segmentation results in small gaps and missing boundaries in the predicted grain boundary map. By incorporating topological information during model training, the connectivity of the grain boundary network and segmentation performance is improved. Finally, our approach is validated by accurate computation on downstream tasks of predicting the underlying grain morphology distributions which are the ultimate quantities of interest for microstructural characterization.

36 MATERIALS SCIENCE↗