Search NASA⌕ Search

SEARCH · Search NASA

Results for “Deep Convolutional 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 379 records · Page 21
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

A physics-informed and hierarchically regularized data-driven model for predicting fluid flow through porous media

This paper presents a new deep learning data-driven model for predicting structure dependent pore-fluid velocity fields in rock. The model is based on a Convolutional Auto-Encoder (CAE) artificial neural network capable of learning from image data generated by direct numerical simulations of fluid flow through pore-structures, such as by Lattice Boltzmann or molecular dynamics methods. The main novelty of the model in comparison to previous CAE-based data-driven approaches consists of three parts. The first is a methodology for decomposing the full-domain of the porous media into sub-regions, or “sub-domains”, in order to reduce the overall size of the CAE, batch process the sub-domains in parallel, and enable the CAE to learn local and generalizable nonlinear mappings of pore-fluid velocities. The second consists of embedding the finite difference solutions of the incompressible Navier-Stokes and continuity equations into convolutional layers prior to the CAE in order to provide the CAE with knowledge of fluid dynamics physics (PhyFlow). The third main novelty is that the training of the CAE is regularized with a hierarchical loss function that encourages the learning of fluid flow patterns (in a way similar to ranked modes in principal component analysis), ranking from most to least important. This is shown to increase the stability in learning, reduce over-fitting, and promote interpretability of the CAE neural network layers (HierCAE). The comprehensive new data-driven model, which we call the PhyFlow-HierCAE model, is shown to exhibit improved accuracy and generalizability of flow field predictions over conventional CAE models, attributable to the embedded physical knowledge and the hierarchical regularization, as well as realize orders of magnitude speed-ups in computation times as a surrogate for the direct numerical simulations. Examples of training and forward predictions on unseen pore-structures are provided and evaluated for data from Lattice Boltzmann and molecular dynamics simulations of pore-fluid flow. The model is shown to be a fast and accurate emulator (or “surrogate”) for predicting effective permeability of unseen pore-structures based on learning from relatively small direct numerical simulation datasets.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

Pre-trained network-based transfer learning: A small-sample machine learning approach to nuclear power plant classification problem

Some research topics belonging to classification problems in the nuclear industry, such as fault diagnosis and accident identification, can be solved by feature extraction and subsequent application of statistical machine learning classifiers. Recently, deep neural network-based methods with automatic feature extraction and high accuracy have gained wide attention. They usually require large-scale training data, however, plant fault or accident data are scarce or difficult to obtain. Here this paper proposes a convolutional network (CNN)-based transfer learning method to solve this problem. The network's shallow layer is derived from a pre-trained CNN based on the ImageNet database to automatically extract features, and the deep layer is customized to match the classification problem. Data in non-image formats are converted to image formats and subsequently used to train the network. Case studies of rotating machines fault diagnosis show that the proposed method requires only limited training data to achieve high accuracy.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2022

Automatic Generation of High-Performance Convolution Kernels on ARM CPUs for Deep Learning

In this work, we present FastConv, a template-based code auto-generation open source library that can automatically generate high-performance deep learning convolution kernels of arbitrary matrices/tensors shapes. FastConv is based on the Winograd algorithm, which is reportedly the highest performing algorithm for the time-consuming convolution layers of convolutional neural networks. ARM CPUs cover a wide range designs and specifications, from embedded devices to HPC-grade CPUs. The leads to the dilemma of how to consistently optimize Winograd-based convolution solvers for convolution layers of different shapes. FastConv addresses this problem by using templates to auto-generate multiple shapes of tuned kernels variants suitable for skinny tall matrices. As a performance portable library, FastConv transparently searches for the best combination of kernel shapes, cache tiles, scheduling of loop orders, packing strategies, access patterns, and online/offline computations. Auto-tuning is used to search the parameter configuration space for the best performance for a given target architecture and problem size. The experiments with layer-wise evaluation on the VGG--16 model confirms a 1.25x performance gains is got by tuning the Winograd library. Integrated comparison results shows 1.02x to 1.40x, 1.14x to 2.17x, and 1.22x and 2.48x speedup is achieved over NNPACK, Arm NN, and FeatherCNN on the Kunpeng 920 beside few cases. Furthermore, problem size performance portability experiments with various convolution shapes shows that FastConv achieves 1.2x to 1.7x speedup and 2x to 22x speedup over NNPACK and ARM NN inference engine using Winograd on Kunpeng 920 . CPU performance portability evaluation on the VGG--16 show an average speedup over NNPACK of 1.42x, 1.21x, 1.26x, 1.37x, 2.26x, and 11.02x is observed on Kunpeng 920, Snapdragon 835, 855, 888, Apple M1, and AWS Graviton2, respectively.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Predicting bulge to total luminosity ratio of galaxies using deep learning

ABSTRACT We present a deep learning model to predict the r-band bulge-to-total luminosity ratio (B/T) of nearby galaxies using their multiband JPEG images alone. Our Convolutional Neural Network (CNN) based regression model is trained on a large sample of galaxies with reliable decomposition into the bulge and disc components. The existing approaches to estimate the B/T ratio use galaxy light-profile modelling to find the best fit. This method is computationally expensive, prohibitively so for large samples of galaxies, and requires a significant amount of human intervention. Machine learning models have the potential to overcome these shortcomings. In our CNN model, for a test set of 20 000 galaxies, 85.7 per cent of the predicted B/T values have absolute error (AE) less than 0.1. We see further improvement to 87.5 per cent if, while testing, we only consider brighter galaxies (with r-band apparent magnitude <17) with no bright neighbours. Our model estimates the B/T ratio for the 20 000 test galaxies in less than a minute. This is a significant improvement in inference time from the conventional fitting pipelines, which manage around 2–3 estimates per minute. Thus, the proposed machine learning approach could potentially save a tremendous amount of time, effort, and computational resources while predicting B/T reliably, particularly in the era of next-generation sky surveys such as the Legacy Survey of Space and Time (LSST) and the Euclid sky survey which will produce extremely large samples of galaxies.

Grover, Harsh (ORCID:0000000321338142)↗
DOE OSTI2023

Extracting the gamma-ray source-count distribution below the Fermi-LAT detection limit with deep learning

We reconstruct the extra-galactic gamma-ray source-count distribution, or dN/dS, of resolved and unresolved sources by adopting machine learning techniques. Specifically, we train a convolutional neural network on synthetic 2-dimensional sky-maps, which are built by varying parameters of underlying source-counts models and incorporate the Fermi-LAT instrumental response functions. The trained neural network is then applied to the Fermi-LAT data, from which we estimate the source count distribution down to flux levels a factor of 50 below the Fermi-LAT threshold. We perform our analysis using 14 years of data collected in the (1,10) GeV energy range. The results we obtain show a source count distribution which, in the resolved regime, is in excellent agreement with the one derived from cataloged sources, and then extends as dN/dS ~ S -2 in the unresolved regime, down to fluxes of 5 · 10 -12 cm -2 s -1 . The neural network architecture and the devised methodology have the flexibility to enable future analyses to study the energy dependence of the source-count distribution.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Using deep machine learning to interpret proton radiography data from a pulsed power experiment

Deep machine learning is used to analyze a proton radiograph from a tin pulsed power experiment and determine density values for each pixel in the image. Two promising convolutional neural network architectures that have proven to be effective for image analysis in other applications are applied to analyze a proton radiograph and find density values. The process of creating a suitable training dataset is described, involving the Lagrangian hydrodynamic model used for simulations of the experiment, the proton radiography forward model to make synthetic images for training, and the manner in which data augmentation is used to expand the resulting image dataset. It is shown that machine learning not only produces a reasonable density field but is also able to predict features in the density field that are suggested by the proton radiograph but not captured by simulations.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Waveform retrieval for ultrafast applications based on convolutional neural networks

Electric field waveforms of light carry rich information about dynamical events on a broad range of timescales. The insight that can be reached from their analysis, however, depends on the accuracy of retrieval from noisy data. In this article, we present a novel approach for waveform retrieval based on supervised deep learning. We demonstrate the performance of our model by comparison with conventional denoising approaches, including wavelet transform and Wiener filtering. The model leverages the enhanced precision obtained from the nonlinearity of deep learning. The results open a path toward an improved understanding of physical and chemical phenomena in field-resolved spectroscopy.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Deep Learning with Reflection High-Energy Electron Diffraction Images to Predict Cation Ratio in Sr 2 x Ti 2(1– x ) O 3 Thin Films

Machine learning (ML) with in-situ diagnostics offers a transformative approach to accelerate, understand, and control thin film synthesis by uncovering relationships between synthesis conditions and material properties. In this study, we demonstrate the application of deep learning to predict the stoichiometry of Sr 2x Ti 2(1–x) O 3 thin films using reflection high-energy electron diffraction images acquired during pulsed laser deposition. A gated convolutional neural network trained for regression of the Sr atomic fraction achieved accurate predictions with a small dataset of 31 samples. Explainable AI techniques revealed a previously unknown correlation between diffraction streak features and cation stoichiometry in Sr 2x Ti 2(1–x) O 3 thin films. Here, our results demonstrate how ML can be used to transform a ubiquitous in-situ diagnostic tool, that is usually limited to qualitative assessments, into a quantitative surrogate measurement of continuously valued thin film properties. Such methods are critically needed to enable real-time control, autonomous workflows, and accelerate traditional synthesis approaches.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2022

A deep-learning based raw waveform region-of-interest finder for the liquid argon time projection chamber

The liquid argon time projection chamber (LArTPC) detector technology has an excellent capability to measure properties of low-energy neutrinos produced by the sun and supernovae and to look for exotic physics at very low energies. In order to achieve those physics goals, it is crucial to identify and reconstruct signals in the waveforms recorded on each TPC wire. In this work, we report on a novel algorithm based on a one-dimensional convolutional neural network (CNN) to look for the region-of-interest (ROI) in raw waveforms. We test this algorithm using data from the ArgoNeuT experiment in conjunction with an improved noise mitigation procedure and a more realistic data-driven noise model for simulated events. This deep-learning ROI finder shows promising performance in extracting small signals and gives an efficiency approximately twice that of the traditional algorithm in the low energy region of ~0.03–0.1 MeV. This method offers great potential to explore low-energy physics using LArTPCs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2021

StressNet - Deep learning to predict stress with fracture propagation in brittle materials

Abstract Catastrophic failure in brittle materials is often due to the rapid growth and coalescence of cracks aided by high internal stresses. Hence, accurate prediction of maximum internal stress is critical to predicting time to failure and improving the fracture resistance and reliability of materials. Existing high-fidelity methods, such as the Finite-Discrete Element Model (FDEM), are limited by their high computational cost. Therefore, to reduce computational cost while preserving accuracy, a deep learning model, StressNet, is proposed to predict the entire sequence of maximum internal stress based on fracture propagation and the initial stress data. More specifically, the Temporal Independent Convolutional Neural Network (TI-CNN) is designed to capture the spatial features of fractures like fracture path and spall regions, and the Bidirectional Long Short-term Memory (Bi-LSTM) Network is adapted to capture the temporal features. By fusing these features, the evolution in time of the maximum internal stress can be accurately predicted. Moreover, an adaptive loss function is designed by dynamically integrating the Mean Squared Error (MSE) and the Mean Absolute Percentage Error (MAPE), to reflect the fluctuations in maximum internal stress. After training, the proposed model is able to compute accurate multi-step predictions of maximum internal stress in approximately 20 seconds, as compared to the FDEM run time of 4 h, with an average MAPE of 2% relative to test data.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Neutrino Physics with Deep Learning on NOvA

The NOvA experiment has made both νμ \nu_\mu disappearance and νe \nu_e appearance measurements in Fermilab's NuMI beam, and is working on cross section measurements using near detector data. At the core of NOvA's measurements is the use of deep learning algorithms for identification and reconstruction of the neutrino flavor and energy. These algorithms, used for the first time on NOvA in 2016, yielded large improvements in selection efficiency, and will be applied to our first anti-neutrino results to be released this year. Presented here is the extension of our deep learning efforts for identification of neutrino signal events, final state identification, single particle tagging, and reconstruction using instance segmentation techniques. We will describe the new implementations of modified Convolutional Neural Networks for anti-neutrino events, single particles and their performance for analysis final states selection, standard candle measurements, and reconstruction.

Psihas, Fernanda [Indiana U.]↗
DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Pushing automated morphological classifications to their limits with the Dark Energy Survey

ABSTRACT We present morphological classifications of ∼27 million galaxies from the Dark Energy Survey (DES) Data Release 1 (DR1) using a supervised deep learning algorithm. The classification scheme separates: (a) early-type galaxies (ETGs) from late-type galaxies (LTGs); and (b) face-on galaxies from edge-on. Our convolutional neural networks (CNNs) are trained on a small subset of DES objects with previously known classifications. These typically have mr ≲ 17.7 mag; we model fainter objects to mr < 21.5 mag by simulating what the brighter objects with well-determined classifications would look like if they were at higher redshifts. The CNNs reach 97 per cent accuracy to mr < 21.5 on their training sets, suggesting that they are able to recover features more accurately than the human eye. We then used the trained CNNs to classify the vast majority of the other DES images. The final catalogue comprises five independent CNN predictions for each classification scheme, helping to determine if the CNN predictions are robust or not. We obtain secure classifications for ∼87 per cent and 73 per cent of the catalogue for the ETG versus LTG and edge-on versus face-on models, respectively. Combining the two classifications (a) and (b) helps to increase the purity of the ETG sample and to identify edge-on lenticular galaxies (as ETGs with high ellipticity). Where a comparison is possible, our classifications correlate very well with Sérsic index (n), ellipticity (ϵ), and spectral type, even for the fainter galaxies. This is the largest multiband catalogue of automated galaxy morphologies to date.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Discovering faint and high apparent motion rate near-Earth asteroids using a deep learning program

ABSTRACT Although many near-Earth objects have been found by ground-based telescopes, some fast-moving ones, especially those near detection limits, have been missed by observatories. We developed a convolutional neural network for detecting faint fast-moving near-Earth objects. It was trained with artificial streaks generated from simulations and was able to find these asteroid streaks with an accuracy of 98.7 per cent and a false positive rate of 0.02 per cent on simulated data. This program was used to search image data from the Zwicky Transient Facility (ZTF) in four nights in 2019, and it identified six previously undiscovered asteroids. The visual magnitudes of our detections range from ∼19.0 to 20.3 and motion rates range from ∼6.8 to 24 deg d−1, which is very faint compared to other ZTF detections moving at similar motion rates. Our asteroids are also ∼1–51 m diameter in size and ∼5–60 lunar distances away at close approach, assuming their albedo values follow the albedo distribution function of known asteroids. The use of a purely simulated data set to train our model enables the program to gain sensitivity in detecting faint and fast-moving objects while still being able to recover nearly all discoveries made by previously designed neural networks which used real detections to train neural networks. Our approach can be adopted by any observatory for detecting fast-moving asteroid streaks.

Wang, Franklin (ORCID:0000000344887355)↗
DOE OSTI2020

Using Deep Learning to Derive Shear-Wave Velocity Models from Surface-Wave Dispersion Data

We present a new algorithm for derivations of 1D shear-wave velocity models from surface-wave dispersion data using convolutional neural networks (CNNs). The technique is applied for continental China and the plate boundary region in southern California. Different CNNs are designed for these two regions and are trained using theoretical Rayleigh-wave phase and group velocity images computed from reference 1D VS models. The methodology is tested with 3260 phase–group images for continental China and 4160 phase–group images for southern California. The conversions of these images to velocity profiles take ~23 s for continental China and ~30 s for southern California on a personal laptop with the NVIDIA GeForce GTX 1060 core and a memory of 6 GB. The results obtained by the CNNs show high correlation with previous studies using conventional methods. The effectiveness of the CNN technique makes this fast method an important alternative for deriving shear-wave velocity models from large datasets of surface-wave dispersion data.

Geochemistry & Geophysics↗
DOE OSTI2023

When Spectral Modeling Meets Convolutional Networks: A Method for Discovering Reionization-era Lensed Quasars in Multiband Imaging Data

Over the last two decades, around 300 quasars have been discovered at z ≳ 6, yet only one has been identified as being strongly gravitationally lensed. We explore a new approach—enlarging the permitted spectral parameter space, while introducing a new spatial geometry veto criterion—which is implemented via image-based deep learning. We first apply this approach to a systematic search for reionization-era lensed quasars, using data from the Dark Energy Survey, the Visible and Infrared Survey Telescope for Astronomy Hemisphere Survey, and the Wide-field Infrared Survey Explorer. Our search method consists of two main parts: (i) the preselection of the candidates, based on their spectral energy distributions (SEDs), using catalog-level photometry; and (ii) relative probability calculations of the candidates being a lens or some contaminant, utilizing a convolutional neural network (CNN) classification. The training data sets are constructed by painting deflected point-source lights over actual galaxy images, to generate realistic galaxy–quasar lens models, optimized to find systems with small image separations, i.e., Einstein radii of θ E ≤ 1''. Visual inspection is then performed for sources with CNN scores of P lens > 0.1, which leads us to obtain 36 newly selected lens candidates, which are awaiting spectroscopic confirmation. These findings show that automated SED modeling and deep learning pipelines, supported by modest human input, are a promising route for detecting strong lenses from large catalogs, which can overcome the veto limitations of primarily dropout-based SED selection approaches.

High-redshift galaxies↗
DOE OSTI2021

Fully Convolutional Spatio-Temporal Models for Representation Learning in Plasma Science

We have trained a fully convolutional spatio-temporal model for fast and accurate representation learning in the challenging exemplar application area of fusion energy plasma science. The onset of major disruptions is a critically important fusion energy science issue that must be resolved for advanced tokamak plasmas such as the $25B burning plasma international thermonuclear experimental reactor (ITER) experiment. While a variety of statistical methods have been used to address the problem of tokamak disruption prediction and control, recent approaches based on deep learning have proven particularly compelling. In the present paper, we introduce further improvements to the fusion recurrent neural network (FRNN) software suite, which delivered cross-machine disruption predictions with unprecedented accuracy using a large database of experimental signals from two major tokamaks. Up to now, FRNN was based on the long short-term memory (LSTM) variant of recurrent neural networks to leverage the temporal information in the data. Here, we implement and apply the "temporal convolutional neural network (TCN)" architecture to the time-dependent input signals. Furthermore, this allows highly optimized convolution operations to carry the majority of the computational load of training, thus enabling a reduction in training time, and the effective use of high-performance computing resources for hyperparameter tuning. At the same time, the TCN-based architecture achieves better predictive performance when compared with the LSTM architecture for various tasks for a representative fusion database.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Power System Event Identification Based on Deep Neural Network With Information Loading

Online power system event identification and classification are crucial to enhancing the reliability of transmission systems. In this study, we develop a deep neural network (DNN) based approach to identify and classify power system events by leveraging real-world measurements from hundreds of phasor measurement units (PMUs) and labels from thousands of events. Two innovative designs are embedded into the baseline model built on convolutional neural networks (CNNs) to improve the event classification accuracy. First, we propose a graph signal processing based PMU sorting algorithm to improve the learning efficiency of CNNs. Second, we deploy information loading based regularization to strike the right balance between memorization and generalization for the DNN. Numerical results based on real-world dataset from the Eastern Interconnection of the U.S power transmission grid show that the combination of PMU based sorting and the information loading based regularization techniques help the proposed DNN approach achieve highly accurate event identification and classification results.

24 POWER TRANSMISSION AND DISTRIBUTION↗