550909 research outputs found
Sort by
Search Tree Pruning for Progressive Neural Architecture Search
Get PDFOur neural architecture search algorithm progressively searches a tree of neural network architectures. Child nodes are created by inserting new layers determined by a transition graph into a parent network up to a maximum depth and pruned when performance is worse than its parent. This increases efficiency but makes the algorithm greedy. Simpler networks are successfully found before more complex ones that can achieve benchmark performance similar to other top-performing networks
Biases in the In Situ Measurement of Particulate Organic Carbon and Its Effect on the Calibration and Validation of Ocean Color Sensors
Get PDFParticulate organic carbon (POC) plays an oversize role, relative to its standing stock in the global carbon (C) cycle. Accurate measurement of POC is central to understanding the ocean C flux and its sensitivity to climate forcing. POC is a standard NASA ocean color data product, which lacks a consensus, quality-assured measurement protocol for satellite validation. Thus, algorithms based on field measurements lacking verified uncertainties have limited applicability towards climate data records. Different sampling and filtration protocols, and blank corrections, introduce biases in the magnitude of POC measured from the field. A significant filter blank attributable to dissolved organic C (DOC) adsorption that, until recently has been seldom corrected for, likely has introduced biases in POC global datasets
Nutrient Addition Effects on Phytoplankton Communities in the Amazon River Plume
Get PDFThe types and abundance of phytoplankton is largely controlled by availability of sunlight and bioavailable nutrients. Phytoplankton require essential nutrients including nitrate, phosphate, and silicate to grow, so understanding the role of these macronutrients in limiting the growth phytoplankton communitiesand the way this may differ depending on community compositionis key to understanding the controls on phytoplankton biomass and community structure. We aimed to explore how the availability of these nutrients affects the health and composition of phytoplankton communities by conducting a series of nutrient amendment experiments (NAEs) with samples from the Western Tropical North Atlantic, which is heavily influenced by the nutrient-rich, low salinity waters of the Amazon River Plume. These experiments, conducted at five locations in and around the plume, provide greater resolution and further our understanding about the ways nutrients affect communities in dynamic coastal regions
Application of Machine Learning Techniques to Aviation Operations: NASA Case Studies
Get PDFThere is an increasing interest in applying methods based on Machine Learning Techniques(MLT) to problems in aviation operations. The current interest is based on developments in Cloud Computing, the availability of open software and the success of MLT in automation, consumer behavior and finance involving large database. Historically aviation operations have been analyzed using physics-based models and provide information for making operational decisions. This talk describes issues to be addressed in applying either model-driven or data-driven methods. Aviation operations involving many decision makers, multiple objectives, poor or unavailable physics-based models and a rich historical database are prime candidates for analysis using data-driven methods. The issues relating to data, feature selection and validation of the models are illustrated by examining case studies of the application of MLT to problems in air traffic management at NASA. Further research is needed in the application of MLT to critical aviation operations. As always, the best approach depends on the task, the physical understanding of the problem and the quality and quantity of the available data
Computationally Efficient Modeling and Data Assimilation of Near-Surface Variability
Get PDFNear-surface (< 20m) ocean exhibits high variability due to coupled interactions, for e.g., with the atmosphere, sea ice, land, etc. Here we focus on atmospheric heat and momentum (wind) forcing, which are known to cause diurnal variability within the mixed layer. Only recently with a combination of sufficiently high vertical/horizontal resolution (75L, 1/4deg) and sub-daily atmospheric forcing fields, ocean models are starting to resolve this diurnal variability. However, the computation expense of such a high vertical resolution is burdensome in the context of coupled modeling and data assimilation. An alternative approach is to parameterize this diurnal variability with a prognostic model, that is embedded into the ocean model.In the first part of this presentation, we will demonstrate results with the above two approaches, by comparing them to profiles of near-surface temperature and salinity. In the context of data assimilation and reanalysis, this modeling capability opens the door to re-examine and perhaps improve specification of background (or, ensemble) error characteristics. The second half of this talk will focus on illustrating diurnally varying errors within an ensemble DA, and possible approaches to improve localization (horizontal/vertical) to extract maximum possible observational information content from in-situ and satellite observations of sea surface temperature
Aircraft Analysis Using the Layered and Extensible Aircraft Performance System (LEAPS)
Get PDFThe Layered and Extensible Aircraft Performance System (LEAPS) is a new air- craft analysis tool being developed by members of the Aeronautics Systems Analysis Branch (ASAB) and the Vehicle Analysis Branch (VAB) at NASA Langley Research Center. LEAPS will enable the analysis of advanced aircraft concepts and architec- tures that include electric and hybrid-electric propulsion systems. The development of LEAPS is motivated by the analysis gaps found in traditional aircraft analysis tools such as the Flight Optimization System (FLOPS). FLOPS has been the tool of choice of the ASAB for over 30 years and has proven to be a reliable analysis tool for conventional aircraft. However, FLOPS is not suitable to analyze the cur- rent unconventional vehicles that are of interest to industry, government agencies, and academia. In contrast, LEAPS is being developed with a flexible architecture that leverages new analysis methodologies that will enable the analysis of unconven- tional aircraft. This paper presents the first complete working version of LEAPS by showing the analysi at include fuel-based and hybrid-electric conceptual aircraft
Prospectives in Deep Space Infrastructures, Development, and Colonization
Get PDFThe realization of the long studied cost reduction benefits of reusable rockets is expected to revolutionize and enable both commercial deep space beyond Geostationary Earth Orbit (GEO) and solar system human colonization. The projections for a myriad of space commercialization activities beyond the current largely positional Earth utilities and Humans Mars both safe and affordable may now be realizable. This report considers these putative commercial and colonizationrelated activities, the emerging technologies, the space functionalities to support and further enable them, and envisions the nature of space developments beyond GEO going forward
Honeywell Enhancing Airplane State Awareness (EASA) Project: Final Report on Refinement and Evaluation of Candidate Solutions for Airplane System State Awareness
Get PDFThe loss of pilot airplane state awareness (ASA) has been implicated as a factor in several aviation accidents identified by the Commercial Aviation Safety Team (CAST). These accidents were investigated to identify precursors to the loss of ASA and develop technologies to address the loss of ASA. Based on a gap analysis, two technologies were prototyped and assessed with a formative pilot-in-the-loop evaluation in NASA Langleys full-motion Research Flight Deck. The technologies address: 1) data source anomaly detection in real-time, and 2) intelligent monitoring aids to provide nominal and predictive awareness of situations to be monitored and a mission timeline to visualize events of interest. The evaluation results indicated favorable impressions of both technologies for mitigating the loss of ASA in terms of operational utility, workload, acceptability, complexity, and usability. The team concludes that there is a feasible retrofit solution for improving ASA that would minimize certification risk, integration costs, and training impact
Fiber-Optic Strain-Based Deflection and Twist Sensing for a High-Aspect-Ratio Swept Wing
Get PDFDesigns of aircraft structures have been moving toward leaner, lightweight designs for increased fuel efficiency. The Passive Aeroelastic Tailored (PAT) wing developed under the NASA Advanced Air Transport Technology (AATT) project is an example of a swept-wing design with high aspect ratio that incorporates lightweight highly-flexible tailored composite construction. The passive aeroelastic tailored structural design has explored the design space to enable aeroelastically tailored wing structures to increase aspect ratios (from 9 to 14) and ultimately reduce weight by 20 percent to 25 percent without impacting aeroelastic performance. To further study the aeroelastic performance of such a wing, the NASA Armstrong Flight Research Center (AFRC) (Edwards, California) has developed efficient real-time structural algorithms that are used in conjunction with a fiber-optic measurement system for lightweight vehicle applications. The AFRC Fiber Optic Sensing System (FOSS) provides up to 8,000 distributed surface strain measurements at one-half-inch increments and can be used to estimate a variety of structural parameters such as shape and load. This report discusses the implementation of strain-based displacement and twist-sensing techniques applied to the PAT wing test article tested at the NASA AFRC Flight Loads Laboratory. Empirical FOSS strain data are collected under varying loading conditions. Strain data are processed with the displacement and twist-sensing algorithms and independently verified by comparison to conventional ground-based instrumentation
