Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep 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.

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

At least 271 records · Page 15
All sourcesNASA NTRSDOE OSTI
NASA NTRS2018

Radio Science System Design and Measurement Results for the NASA Deep Space Network (DSN)

Radio science measurements have been performed using the NASA Deep Space Network (DSN) with many different spacecraft over several decades. Radio science has been used for the study of planetary atmospheres, the solar corona and the search for gravity waves, among other things. The majority of these measurements are made using the X and Ka-band deep space bands. Although the primary mission for the DSN is tracking, telemetry and command (TT&C) for NASA’s many deep-space spacecraft, radio science measurements continue to be an important secondary mission. The science requirements for these measurements have resulted in stringent performance requirements for both the spacecraft and ground system equipment. In particular, the requirements for amplitude stability, phase stability (Allen deviation) and phase noise are very demanding. The system Allen deviation requirement at Ka-band is < 2.4 E-15 over 1000 seconds, while the phase noise requirement is < -50 dBc/Hz for a 1 Hz offset. Various design techniques have been used for the DSN radio frequency (RF) electronics, high power transmitters and antenna structures to meet the stringent requirements for all 3 of these parameters. Some details for the design techniques will be described in the paper. Another important consideration for a radio science system is the verification approach for components, as well as for individual subsystems and then the overall system. Phase-locked oscillators (PLOs) are one of the key component types that determine overall phase noise and Allen deviation system performance. Measurement techniques used for PLOs, as well as for the overall ground system, will be discussed. Measurement results for the 2 new DSN antennas, recently built under the DSN Aperture Enhancement Project (DAEP) will also be shown. In addition, some recent radio science measurements from the Cassini and JUNO missions, using the new antennas, will be presented.

LaBelle, Remi C.↗
DOE OSTI2023

Improving Deep Neural Networks’ Training for Image Classification With Nonlinear Conjugate Gradient-Style Adaptive Momentum

Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant computational burden. In this article, we propose a novel adaptive momentum for improving DNNs training; this adaptive momentum, with no momentum-related hyperparame- ter required, is motivated by the nonlinear conjugate gradient (NCG) method. Stochastic gradient descent (SGD) with this new adaptive momentum eliminates the need for the momentum hyperparameter calibration, allows using a significantly larger learning rate, accelerates DNN training, and improves the final accuracy and robustness of the trained DNNs. For example, SGD with this adaptive momentum reduces classification errors for training ResNet110 for CIFAR10 and CIFAR100 from 5.25% to 4.64% and 23.75% to 20.03%, respectively. Furthermore, SGD, with the new adaptive momentum, also benefits adversarial training and, hence, improves the adversarial robustness of the trained DNNs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Deep operator network surrogate for phase-field modeling of metal grain growth during solidification

A deep operator network (DeepONet) has been constructed that generates accurate representations of phase-field model simulations for evolving two dimensional metal grain morphology growing from melt. These representations serve as lower resolution, computationally efficient stand-ins for quick parameter space exploration of solutions to the the Allen-Cahn equations that dictate the phase-field model simulations. The experimental target for the phase-field model is a uranium casting system cooling a 434 g uranium charge from a maximum temperature of 1400° C at an average rate of 30° C / min , traversing the crystallographic phases of the pure metal. Experimental parameters inform the phase-field model, whose higher resolution computational model solutions are used to train the DeepONet in a given parameter space with the aim of developing a faster, more efficient method for predicting the solidifying metal's microstructure at different potential experimental values. The final DeepONet generates high accuracy, lower resolution predictions with cumulative relative approximation error over all timesteps of less than 0.5%, while ensuring solutions remain within physically feasible ranges. Further, these relative error values are comparable with other state-of-the-art DeepONet models for microstructure evolution, while significantly reducing the amount of training data required. Training a convolutional neural network simultaneously with the DeepONet, enforcing realistic values at the complex metal grain boundaries, and mathematically encoding boundary conditions into the structure of the DeepONet improved prediction accuracy and computational efficiency over a standard DeepONet model.

36 MATERIALS SCIENCE↗
NASA NTRS2012

Frequency Domain Beamforming for a Deep Space Network Downlink Array

This paper describes a frequency domain beamformer to array up to 8 antennas of NASA's Deep Space Network currently in development. The objective of this array is to replace and enhance the capability of the DSN 70m antennas with multiple 34m antennas for telemetry, navigation and radio science use. The array will coherently combine the entire 500 MHz of usable bandwidth available to DSN receivers. A frequency domain beamforming architecture was chosen over a time domain based architecture to handle the large signal bandwidth and efficiently perform delay and phase calibration. The antennas of the DSN are spaced far enough apart that random atmospheric and phase variations between antennas need to be calibrated out on an ongoing basis in real-time. The calibration is done using measurements obtained from a correlator. This DSN Downlink Array expands upon a proof of concept breadboard array built previously to develop the technology and will become an operational asset of the Deep Space Network. Design parameters for frequency channelization, array calibration and delay corrections will be presented as well a method to efficiently calibrate the array for both wide and narrow bandwidth telemetry.

Filterbank↗
DOE OSTI2025

An adaptive and stability-promoting layerwise training approach for sparse deep neural network architecture

This work presents a two-stage adaptive framework for progressively developing deep neural network (DNN) architectures that generalize well for a given training data set. In the first stage, a layerwise training approach is adopted where a new layer is added each time and trained independently by freezing parameters in the previous layers. We impose desirable structures on the DNN by employing manifold regularization, sparsity regularization, and physics-informed terms. We introduce a ε – δ – stability-promoting concept as a desirable property for a learning algorithm and show that employing manifold regularization yields a ε – δ stability-promoting algorithm. Further, we also derive the necessary conditions for the trainability of a newly added layer and investigate the training saturation problem. In the second stage of the algorithm (post-processing), a sequence of shallow networks is employed to extract information from the residual produced in the first stage, thereby improving the prediction accuracy. Numerical investigations on prototype regression and classification problems demonstrate that the proposed approach can outperform fully connected DNNs of the same size. Moreover, by equipping the physics-informed neural network (PINN) with the proposed adaptive architecture strategy to solve partial differential equations, we numerically show that adaptive PINNs not only are superior to standard PINNs but also produce interpretable hidden layers with provable stability. As a result, we also apply our architecture design strategy to solve inverse problems governed by elliptic partial differential equations.

42 ENGINEERING↗
NASA NTRS1989

Automated monitor and control for deep space network subsystems

The problem of automating monitor and control loops for Deep Space Network (DSN) subsystems is considered and an overview of currently available automation techniques is given. The use of standard numerical models, knowledge-based systems, and neural networks is considered. It is argued that none of these techniques alone possess sufficient generality to deal with the demands imposed by the DSN environment. However, it is shown that schemes that integrate the better aspects of each approach and are referenced to a formal system model show considerable promise, although such an integrated technology is not yet available for implementation. Frequent reference is made to the receiver subsystem since this work was largely motivated by experience in developing an automated monitor and control loop for the advanced receiver.

Smyth, P.↗
NASA NTRS1991

Overview of arraying techniques in the deep space network

Four different arraying schemes that can be used by the Deep Space Network are functionally discussed and compared. These include symbol stream combining (SSC), baseband combining (BC), carrier arraying (CA), and full spectrum combining (FSC). In addition, sibeband aiding (SA) is also included and compared even though it is not an arraying scheme, since it uses a single antenna. Moreover, combinations of these schemes are discussed, such as carrier arraying with sideband aiding and baseband combining (CA/SA/BC) or carrier arraying with symbol stream combining (CA/SSC). Complexity versus performance is traded off and the benefits to the reception of existing spacecraft signals are discussed. Recommendations are made as to the best techniques for particular configurations.

Mileant, A.↗
NASA NTRS2006

Identifying Trends in Deep Space Network Monitor Data

A computer program has been developed that analyzes Deep Space Network monitor data, looking for changes of trends in critical parameters. This program represents a significant improvement over the previous practice of manually plotting data and visually inspecting the resulting graphs to identify trends. This program uses proven numerical techniques to identify trends. When a statistically significant trend is detected, then it is characterized by means of a symbol that can be used by pre-existing model-based reasoning software. The program can perform any of the following functions: Given an expectation that data in a given list should exhibit an upward, downward, constant, or unknown trend, it can determine whether the data do or do not follow such a trend. Given a list of data, it can identify which of the aforementioned trends the data follow. Given two lists of data, it can determine whether or not both follow the same trend. This program can be executed on a variety of computers. It can be distributed in either source code or binary code form. It must be run in conjunction with any one of a number of Lisp compilers that are available commercially or as shareware.

James, Mark↗
NASA NTRS2022

Differential Ranging Experiments with NASA’s High-Gain Deep Space Network Antennas

Radar experiments are currently under way at the Goldstone Deep Space Communications Complex (GDSCC) to determine the range of cis-lunar objects using a single high-gain 34m antenna transmitting 20 kW at X-band wavelengths, and two 34m receivers located at the GDSCC. The 34m antennas of the Deep Space Network (DSN) have effective fields-of-view of approximately 1 milli-radians, hence predetermined pointing predicts with accuracies on the order of 1 milli-degree are required. The extremely accurate 1 nano-second timing distribution of the DSN network enables sub-meter level ranging in the radial direction, by measuring delays in phase-modulated pseudonoise (PN) echoes reflected from the target. In this paper we demonstrate experimental verification of the proposed techniques by post-processing radar reflections obtained from a decommissioned geo-synchronous satellite. The characteristics of correlation-based delay measurements obtained simultaneously with two different receiving antennas are described. These are the fundamental quantities used to form the “time difference of arrival” (TDOA) observables commonly used in GPS positioning algorithms, and the results interpreted in the context of hyperbolic position location that could be used in the future to monitor spacecraft in cis-lunar space.

Cheung, Kar-Ming↗
NASA NTRSDate not supplied

Lunar Navigation Performance Using the Deep Space Network and Alternate Solutions to Support Precision Landing

As human exploration once again targets the surface of the Moon, questions continue to emerge regarding the necessity of Earth-based tracking systems, such as the Deep Space Network, for spacecraft navigation in support of lunar descent and landing. This paper will derive an extensive Deep Space Network sensor model for use in linear covariance analysis and demonstrate the resulting trajectory dispersions and navigation performance in comparison with alternate solutions, such as terrain relative navigation. An in-depth trade study with considerations for various trajectory profiles, time allocated to ground tracking, number of active ground stations, and interaction with other sensors will be conducted to shed significant insight into sensor suite requirements to ensure safe and precise landing on the Moon.

Navigation↗
DOE OSTI2020

DeepMerge: Classifying high-redshift merging galaxies with deep neural networks

In this work, we investigate and demonstrate the use of convolutional neural networks (CNNs) for the task of distinguishing between merging and non-merging galaxies in simulated images, and for the first time at high redshifts (i.e. $z=2$). We extract images of merging and non-merging galaxies from the Illustris-1 cosmological simulation and apply observational and experimental noise that mimics that from the Hubble Space Telescope; the data without noise form a "pristine" data set and that with noise form a "noisy" data set. The test set classification accuracy of the CNN is $79\%$ for pristine and $76\%$ for noisy. The CNN outperforms a Random Forest classifier, which was shown to be superior to conventional one- or two-dimensional statistical methods (Concentration, Asymmetry, the Gini, $M_{20}$ statistics etc.), which are commonly used when classifying merging galaxies. We also investigate the selection effects of the classifier with respect to merger state and star formation rate, finding no bias. Finally, we extract Grad-CAMs (Gradient-weighted Class Activation Mapping) from the results to further assess and interrogate the fidelity of the classification model.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2021

Off-the-shelf deep learning is not enough, and requires parsimony, Bayesianity, and causality

Abstract Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems in speech recognition, computer vision, finance, etc. However, adoption of deep learning in physical domains brings substantial challenges stemming from the correlative nature of deep learning methods compared to the causal, hypothesis driven nature of modern science. We argue that the broad adoption of Bayesian methods incorporating prior knowledge, development of solutions with incorporated physical constraints and parsimonious structural descriptors and generative models, and ultimately adoption of causal models, offers a path forward for fundamental and applied research.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1973

The deep space network, volume 15

The DSN progress is reported in flight project support, TDA research and technology, network engineering, hardware and software implementation, and operations. Topics discussed include: DSN functions and facilities, planetary flight projects, tracking and ground-based navigation, communications, data processing, network control system, and deep space stations.

Source record↗
NASA NTRS1977

Baseband recording and playback

Analog recordings of spacecraft telemetry signals are made by Deep Space Network Deep Space Stations to provide backup for both spacecraft and ground station anomalies. The requirements that must be met to insure successful baseband recording and playback are given. Recording and playback procedures are developed to insure that these requirements are met and performance results from tests conducted at the Compatibility Test Area (JPL) are tabulated.

Kent, S. S.↗
NASA NTRS1979

Automatic filament warm-up controller

As part of the unattended operations objective of the Deep Space Network deep space stations, this filament controller serves as a step between manual operation of the station and complete computer control. Formerly, the operator was required to devote five to fifteen minutes of his time just to properly warm up the filaments on the klystrons of the high power transmitters. The filament controller reduces the operator's duty to a one-step command and is future-compatible with various forms of computer control.

Mccluskey, J.↗