Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep data”

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 181 records · Page 10

Deep Learning Vetting of TESS FFI Data: Results and Comparison with 2-Min Data

We present the results of vetting TCEs from the TESS SPOC full-frame images (FFI) Year 5 data using our deep learning model, and we compare the performance in this dataset against the results obtained for the TESS SPOC 2-min data. The 200-second cadence FFI data expands the search to a list of targets that not only includes 2-minute targets, but also potentially high-value targets within 100 parsecs or with H-magnitude <10, and field targets with TESS magnitude <13.5. This work aims to explore this rich dataset and increase the efficiency and throughput of the vetting process by helping unearth more high-quality planet candidates from the TESS mission.

tess spoc↗

Implementing an Applied Science Program

The work implied in the NASA Applied Science Program requires a delicate balancing act for the those doing it. At the implementation level there are multiple tensions intrinsic to the program. For example each application of an existing product to a decision support process requires deep knowledge about the data and deep knowledge about the decision making process. It is highly probable no one person has this range of knowledge. Otherwise the decision making process would already be using the data. Therefore, a team is required. But building a team usually requires time, especially across agencies. Yet the program mandates efforts of relatively short duration. Further, those who know the data are scientists, which makes them essential to the program. But scientists are evaluated on their publication record. Anything which diverts a scientist from the research for his next publication is an anathema to him and potential death to their career. Trying to get another agency to use NASA data does not strike most scientists as material inherently suitable for publication. Also, NASA wishes to rapidly implement often substantial changes to another agency's process. For many reasons, such as budget and program constraints, speed is important. But the owner of a decision making process is tightly constrained, usually by law, regulation, organization and custom. Changes when made are slow, cautious, even hesitant, and always done according a process specific to the situation. To manage this work MSFC must balance these and other tensions. Some things we have relatively little control over, such as budget. These we try to handle by structural techniques. For example by insisting all of our people work on multiple projects simultaneously we inherently have diversification of funding for all of our people. In many cases we explicitly use some elements of tension to be productive. For example the need for the scientists to constantly publish is motivation to keep tasks short and to the immediate point. As another example, the DSS agency and it personnel are treated as integral to the effort at every step. Thus, their ways of doing things, their problems and their assets become part of the solution. MSFC also manages the Applied Science work' within a strategic framework.' First, the scientists are necessarily at the core of all of this work as well as all of the work within the larger' organization. We therefore strive to keep a roughly 50-50 balance between the work done on tasks funded directly by the research side of NASA's Science Mission Directorate and funding from the applied side. Done at both the organization level and the level of the individual, this keeps the scientist both happy and productive over the long term. We also try diligently to remove as much burden from the scientist as practical by employing people such as the authors and others to do tasks not requiring; scientific knowledge. We also have designed our effort to take full use of external partners. We actively seek and support, including fund, people from multiple organizations to join us as committed collaborators. In this we use today's money and today's problem to help us diversify and strengthen for tomorrow. MSFC also considers the Applied Science work holistically. Each element is viewed as a step in a larger process. At a management level we can chose to emphasize or encourage certain areas which service long term goals. Thus, if we think work in a particular area should be developed, we can start with the smaller, less costly elements and grow.

Rickman, Doug↗

BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset

The BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset adds node-level energy consumption data from watt-meters to the primary sweep of the BUTTER - Empirical Deep Learning Dataset. This dataset contains energy consumption and performance data from 63,527 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (number of trainable parameters), 8 network "shapes", and 14 depths on both CPU and GPU hardware collected using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use, and highlights the impact of cache effects. BUTTER-E is intended to be joined with the BUTTER dataset (see "BUTTER - Empirical Deep Learning Dataset on OEDI" resource below) which characterizes the performance of 483k distinct fully connected neural networks but does not include energy measurements.

Array↗

Physics-Guided Deep Learning for Time-Series State Estimation Against False Data Injection Attacks

The modern power grid is a cyber-physical system. While the grid is becoming more intelligent with emerging sensing and communication techniques, new vulnerabilities are introduced and cyber security becomes a major concern. One type of cyber-attacks - False Data Injection Attacks (FDIAs) - exploits the limitations in traditional power system state estimation, and modifies system states without being detected. In this paper, we propose a physics-guided deep learning (PGDL) approach to defend against FDIAs. The PGDL takes real-time measurements as inputs to neural networks, outputs the estimated states, and reconstructs measurements considering power system physics. A deep recurrent neural network - Long Short-Term Memory (LSTM) - is employed to learn the temporal correlations among states. This hybrid learning model leads to a time-series state estimation method to defend against FDIAs. The simulation results using IEEE 14-bus test system demonstrate the accuracy and robustness of the proposed time-series state estimation under FDIAs.

Cyber security, state estimation, physics-guided d↗

High Degree and Order Gravity Fields of the Moon Derived from GRAIL Data

The Gravity Recovery and Interior Laboratory (GRAIL) spacecraft conducted the mapping of the gravity field of the Moon from March 1, 2012 to May 29, 2012. The twin spacecraft acquired highly precise K Band range-rate (KBRR) intersatellite ranging data and Deep Space Network (DSN) data during this prime mission phase from altitudes of 15 to 75 km above the lunar surface over three lunar months. We have processed these data using the NASA GSFC GEODYN orbit determination and geodetic parameter estimation program, and we have determined gravity fields up to degree and order 420 in spherical harmonics. The new gravity solutions show improved correlations with LOLA-derived topography to high degree and order and resolve many lunar features in the geopotential with a resolution of less than 30 km, including for example the central peak of the crater Tycho. We discuss the methodology used for the processing of the GRAIL data, the quality of the orbit determination on the GRAIL satellites and the derivation of the solutions, and their evaluation with independent data, including Lunar Prospector. We show that with these new GRAIL gravity solutions, we can now fit the low altitude, extended mission Lunar Prospector tracking data better than with any previous gravity model that included the LP data.

Lemoine, F. G.↗

Stochastic availability analysis of operational data systems in the Deep Space Network

Existing availability models of standby redundant systems consider only an operator's performance and its interaction with the hardware performance. In the case of operational data systems in the Deep Space Network (DSN), in addition to an operator system interface, a controller reconfigures the system and links a standby unit into the network data path upon failure of the operating unit. A stochastic (Markovian) process technique is used to model and analyze the availability performance and occurrence of degradation due to partial failures are quantitatively incorporated into the model. Exact expressions of the steady state availability and proportion degraded performance measures are derived for the systems under study. The interaction among the hardware, operator, and controller performance parameters and that interaction's effect on data availability are evaluated and illustrated for an operational data processing system.

Issa, T. N.↗

RF and Optical Communications: A Comparison of High Data Rate Returns From Deep Space in the 2020 Timeframe

As NASA proceeds with plans for increased science data return and higher data transfer capacity for science missions, both RF and optical communications are viable candidates for significantly higher-rate communications from deep space to Earth. With the inherent advantages, smaller apertures and larger bandwidths, of optical communications, it is reasonable to expect that at some point in time and combination of increasing distance and data rate, the rapidly emerging optical capabilities would become more advantageous than the more mature and evolving RF techniques. This paper presents a comparison of the burden to a spacecraft by both RF and optical communications systems for data rates of 10, 100, and 1000 Mbps and large distances. Advanced technology for RF and optical communication systems have been considered for projecting capabilities in the 2020 timeframe. For the comparisons drawn, the optical and RF ground terminals were selected to be similar in cost. The RF system selected is composed of forty-five 12-meter antennas, whereas the selected optical system is equivalent to a 10-meter optical telescope. Potential differences in availability are disregarded since the focus of this study is on spacecraft mass and power burden for high-rate mission data, under the assumption that essential communications will be provided by low-rate, high availability RF. For both the RF and optical systems, the required EIRP, for a given data rate and a given distance, was achieved by a design that realized the lowest possible communications subsystem mass (power + aperture) consistent with achieving the lowest technology risk. A key conclusion of this paper is that optical communications has great potential for high data rates and distances of 2.67 AU and beyond, but requires R&D and flight demonstrations to prove out technologies.

Williams, W. Dan↗

A Whale of a Tale: Creating Spacecraft Telemetry Data Analysis Products for the Deep Impact Mission

A description of the Whale product generation utility and its means of analyzing project data for Deep Impact Missions is presented. The topics include: 1) Whale Definition; 2) Whale Overview; 3) Whale Challenges; 4) Network Configuration; 5) Network Diagram; 6) Whale Data Flow: Design Decisions; 7) Whale Data Flow Diagram; 8) Whale Data Flow; 9) Whale Team and Users; 10) Creeping Requirements; 11) Whale Competition; 12) Statistics: Processing Time; 13) CPU and Disk Usage; 14) The Ripple Effect of More Data; and 15) Data Validation and the Automation Challenge.

ground data systems↗

A Whale of a Tale: Creating Spacecraft Telemetry Data Analysis Products for the Deep Impact Mission

This paper describes some of the challenges and lessons learned from the Deep Impact (DI) Mission Ground Data System's (GDS) telemetry data processing and product generation tool, nicknamed 'Whale.' One of the challenges of any mission is to analyze testbed and operational telemetry data. Methods to retrieve this data to date have required spacecraft subsystem members to become experts in the use of a myriad of query and plot tools. As budgets shrink, and the GDS teams grow smaller, more of the burden to understand these tools falls on the users. The user base also varies from novice to expert, and requiring them to become GDS tool experts in addition to spacecraft domain experts is an undue burden. The "Whale" approach is to process all of the data for a given spacecraft test, and provide each subsystem with plots and data products 'automagically.'.

Deep Impact↗

Data for FUN-PROSE: A Deep Learning Approach to Predict Condition-Specific Gene Expression in Fungi

mRNA levels of all genes in a genome is a critical piece of information defining the overall state of the cell in a given environmental condition. Being able to reconstruct such condition-specific expression in fungal genomes is particularly important to metabolically engineer these organisms to produce desired chemicals in industrially scalable conditions. Most previous deep learning approaches focused on predicting the average expression levels of a gene based on its promoter sequence, ignoring its variation across different conditions. Here we present FUN-PROSE—a deep learning model trained to predict differential expression of individual genes across various conditions using their promoter sequences and expression levels of all transcription factors. We train and test our model on three fungal species and get the correlation between predicted and observed condition-specific gene expression as high as 0.85. We then interpret our model to extract promoter sequence motifs responsible for variable expression of individual genes. We also carried out input feature importance analysis to connect individual transcription factors to their gene targets. A sizeable fraction of both sequence motifs and TF-gene interactions learned by our model agree with previously known biological information, while the rest corresponds to either novel biological facts or indirect correlations.

Genomics↗

Optical variability and pigment biomass in the Sargasso Sea as determined using deep-sea optical mooring data

A 9-month time series of optical data, composed of three separate deployments, from the Biowatt 1987 deep-sea mooring located in the oligotrophic waters of the Sargasso Sea at 32 deg N, 70 deg W, is presented. These data were analyzed using several biooptical models for the purpose of providing a continuous estimate of the spectral diffuse attenuation coefficient and pigment biomass. Results of the first 70-day deployment demonstrate the utility of optical observations as proxy measures of biological parameters, and facilitate a preliminary analysis of the temporal variations and a comparison of the present biooptical observations with concurrent physical observations. Biooptical properties exhibit variation in response to physical forcings, such as diel variations of incident solar irradiance, episodic wind forcing, advective mesoscale eddy layer depth, and succession of phytoplankton. The mooring data are discussed in the context of contemporaneous shipboard observations as well as in terms of future ocean satellite observations.

Smith, R. C.↗

A deep learning-accelerated data assimilation and forecasting workflow for commercial-scale geologic carbon storage

Fast assimilation of monitoring data to update forecasts of pressure buildup and carbon dioxide (CO 2 ) plume migration under geologic uncertainties is a challenging problem in geologic carbon storage. The high computational cost of data assimilation with a high-dimensional parameter space impedes fast decision-making for commercial-scale reservoir management. We propose to leverage physical understandings of porous medium flow behavior with deep learning techniques to develop a fast data assimilation-reservoir response forecasting workflow. Applying an Ensemble Smoother Multiple Data Assimilation (ES-MDA) framework, the workflow updates geologic properties and predicts reservoir performance with quantified uncertainty from pressure history and CO 2 plumes interpreted through seismic inversion. As the most computationally expensive component in such a workflow is reservoir simulation, we developed surrogate models to predict dynamic pressure and CO 2 plume extents under multi-well injection. The surrogate models employ deep convolutional neural networks, specifically, a wide residual network and a residual U-Net. The workflow is validated against a flat threedimensional reservoir model representative of a clastic shelf depositional environment. Intelligent treatments are applied to bridge between quantities in a true-3D reservoir model and those in a single-layer reservoir model. The workflow can complete history matching and reservoir forecasting with uncertainty quantification in less than one hour on a mainstream personal workstation.

25 ENERGY STORAGE↗

CO2 Plume Imaging with Accelerated Deep Learning-based Data Assimilation Considering Multiple Realizations: Application to the Illinois Basin-Decatur Carbon Sequestration Project

We propose a fast and efficient deep learning workflow for near real-time data assimilation, forecasting and visualization of CO2 plume evolution in saline aquifer and demonstrate its application at a field site. Unlike the previous work, this study incorporates the impact of spatial heterogeneity using multiple realizations. In the proposed workflow, a neural network model utilizes available monitoring data such as downhole pressure measurements as input and predicts the propagating pressure ‘front’ using the diffusive time of flight (DTOF) map which is considered as representative reservoir image of the flow field. The DTOF is the arrival time of pressure front propagation, which can be computed by the Fast Marching Method rapidly without flow simulations. Reservoir model calibration can be implemented by selecting the training data samples that describe the predicted DTOF map based on observed data. The power and efficacy of our workflow is demonstrated by application to the Illinois Basin-Decatur Project.

CO2 plume imaging↗

Deep learning categorization of infrasound array data

Here we develop a deep learning-based infrasonic detection and categorization methodology that uses convolutional neural networks with self-attention layers to identify stationary and non-stationary signals in infrasound array processing results. Using features extracted from the coherence and direction-of-arrival information from beamforming at different infrasound arrays, our model more reliably detects signals compared with raw waveform data. Using three infrasound stations maintained as part of the International Monitoring System, we construct an analyst-reviewed data set for model training and evaluation. We construct models using a 4-category framework, a generalized noise vs non-noise detection scheme, and a signal-of-interest (SOI) categorization framework that merges short duration stationary and non-stationary categories into a single SOI category. We evaluate these models using a combination of k-fold cross-validation, comparison with an existing “state-of-the-art” detector, and a transportability analysis. Although results are mixed in distinguishing stationary and non-stationary short duration signals, f-scores for the noise vs non-noise and SOI analyses are consistently above 0.96, implying that deep learning-based infrasonic categorization is a highly accurate means of identifying signals-of-interest in infrasonic data records.

47 OTHER INSTRUMENTATION↗