Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network reduction”

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 1,009 records · Page 56
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

Wireless Sensor Networks - Node Localization for Various Industry Problems

Fast, effective monitoring following airborne releases of toxic substances is critical to mitigate risks to threatened population areas. Wireless sensor nodes at fixed predetermined locations may monitor such airborne releases and provide early warnings to the public. A challenging algorithmic problem is determining the locations to place these sensor nodes while meeting several criteria: 1) provide complete coverage of the domain, and 2) create a topology with problem dependent node densities, while 3) minimizing the number of sensor nodes. This manuscript presents a novel approach to determining optimal sensor placement, Advancing Front mEsh generation with Constrained dElaunay Triangulation and Smoothing (AFECETS) that addresses these criteria. A unique aspect of AFECETS is the ability to determine wireless sensor node locations for areas of high interest (hospitals, schools, high population density areas) that require higher density of nodes for monitoring environmental conditions, a feature that is difficult to find in other research work. The AFECETS algorithm was tested on several arbitrary shaped domains. AFECETS simulation results show that the algorithm 1) provides significant reduction in the number of nodes, in some cases over 40%, compared to an advancing front mesh generation algorithm, 2) maintains and improves optimal spacing between nodes, and 3) produces simulation run times suitable for real-time applications.

42 ENGINEERING↗
DOE OSTI2021

The impact of Uber and Lyft on vehicle ownership, fuel economy, and transit across U.S. cities

We estimate the effects of transportation network companies (TNCs) Uber and Lyft on vehicle ownership, fleet average fuel economy, and transit use in U.S. urban areas using a set of difference-in-difference propensity score-weighted regression models that exploit staggered market entry across the U.S. from 2011 to 2017. We find evidence that TNC entry into urban areas causes an average 0.7% increase in vehicle registrations with significant heterogeneity in these effects across urban areas: TNC entry produces larger vehicle ownership increases in urban areas with higher initial ownership (car-dependent cities) and in urban areas with lower population growth (where TNC-induced vehicle adoption outpaces population growth). We also find no statistically significant average effect of TNC entry on fuel economy or transit use but find evidence of heterogeneity in these effects across urban areas, including larger transit ridership reductions after TNC entry in areas with higher income and more childless households.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
NASA NTRS2022

Microstructure Segmentation with Deep Learning Encoders Pre-Trained on a Large Microscopy Dataset

This study examined the improvement of microscopy segmentation accuracy by transfer learning from a large dataset of microscopy images called MicroNet. Many neural network encoder architectures, including VGG, Inception, and ResNet, were trained on over 100,000 labelled microscopy images from 54 classes. These pre-trained encoders were then embedded into multiple segmentation architectures including U-Net and DeepLabV3+ to evaluate segmentation performance on newly created benchmark microscopy datasets. Compared to ImageNet pre-training, models pre-trained on MicroNet generalized better to out-of-distribution micrographs taken under different imaging and sample conditions and were more accurate with less training data. When training with only a single Ni-superalloy image, pre-training on MicroNet produced a 72.2 percent reduction in relative segmentation error. These results suggest that transfer learning from large in-domain datasets generate models with learned feature representations that are more useful for downstream tasks and will likely improve any microscopy image analysis technique that can leverage pre-trained encoders.

machine learning↗
DOE OSTI2019

A Deep Learning Parameterization for Ozone Dry Deposition Velocities

Abstract The loss of ozone to terrestrial and aquatic systems, known as dry deposition, is a highly uncertain process governed by turbulent transport, interfacial chemistry, and plant physiology. We demonstrate the value of using Deep Neural Networks (DNN) in predicting ozone dry deposition velocities. We find that a feedforward DNN trained on observations from a coniferous forest site (Hyytiälä, Finland) can predict hourly ozone dry deposition velocities at a mixed forest site (Harvard Forest, Massachusetts) more accurately than modern theoretical models, with a reduction in the normalized mean bias (0.05 versus ~0.1). The same DNN model, when driven by assimilated meteorology at 2° × 2.5° spatial resolution, outperforms the Wesely scheme as implemented in the GEOS‐Chem model. With more available training data from other climate and ecological zones, this methodology could yield a generalizable DNN suitable for global models.

Silva, S. J.↗
DOE OSTI2020

Rapid Mineral Precipitation During Shear Fracturing of Carbonate-Rich Shales

Target subsurface reservoirs for emerging low-carbon energy technologies and geologic carbon sequestration typically have low permeability and thus rely heavily on fluid transport through natural and induced fracture networks. Sustainable development of these systems requires deeper understanding of how geochemically mediated deformation impacts fracture microstructure and permeability evolution, particularly with respect to geochemical reactions between pore fluids and the host rock. Here, a series of triaxial direct shear experiments was designed to evaluate how fractures generated at subsurface conditions respond to penetration of reactive fluids with a focus on the role of mineral precipitation. Calcite-rich shale cores were directly sheared under 3.5 MPa confining pressure using BaCl 2 -rich solutions as a working fluid. Experiments were conducted within an X-ray computed tomography (xCT) scanner to capture 4-D evolution of fracture geometry and precipitate growth. Three shear tests evidenced nonuniform precipitation of barium carbonates (BaCO 3 ) along through-going fractures, where the extent of precipitation increased with increasing calcite content. Precipitates were strongly localized within fracture networks due to mineral, geochemical, and structural heterogeneities and generally concentrated in smaller apertures where rock:water ratios were highest. The combination of elevated fluid saturation and reactive surface area created in freshly activated fractures drove near-immediate mineral precipitation that led to an 80% permeability reduction and significant flow obstruction in the most reactive core. While most previous studies have focused on mixing-induced precipitation, this work demonstrates that fluid–rock interactions can trigger precipitation-induced permeability alterations that can initiate or mitigate risks associated with subsurface energy systems.

58 GEOSCIENCES↗
DOE OSTI2015

Redesigning rain gauges network in Johor using geostatistics and simulated annealing

Recently, many rainfall network design techniques have been developed, discussed and compared by many researchers. Present day hydrological studies require higher levels of accuracy from collected data. In numerous basins, the rain gauge stations are located without clear scientific understanding. In this study, an attempt is made to redesign rain gauge network for Johor, Malaysia in order to meet the required level of accuracy preset by rainfall data users. The existing network of 84 rain gauges in Johor is optimized and redesigned into a new locations by using rainfall, humidity, solar radiation, temperature and wind speed data collected during the monsoon season (November - February) of 1975 until 2008. This study used the combination of geostatistics method (variance-reduction method) and simulated annealing as the algorithm of optimization during the redesigned proses. The result shows that the new rain gauge location provides minimum value of estimated variance. This shows that the combination of geostatistics method (variance-reduction method) and simulated annealing is successful in the development of the new optimum rain gauge system.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2015

Effect of high pressure on the electrical resistivity of Ge−Te−In glasses

The variation in the electrical resistivity of the chalcogenide glasses Ge{sub 15}Te{sub 85-x}In{sub x} has been studied as a function of high pressure for pressures up to 8.5GPa. All the samples studied undergo a semi-conductor to metallic transition in a continuous manner at pressures between 1.5-2.5GPa. The transition pressure at which the samples turn metallic increases with increase in percentage of Indium. This increase is a direct consequence of the increase in network rigidity with the addition of Indium. At a constant pressure of 0.5GPa, the normalized resistivity shows some signature of the existence of the intermediate phase. Samples recovered after a pressure cycle remain amorphous suggesting that the semi-conductor to metallic transition arises from a reduction of the band gap due to pressure or the movement of the Fermi level into the conduction or valence band.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2023

Preparation for quantum simulation of the ( 1 + 1 ) -dimensional O(3) nonlinear σ model using cold atoms

The 1+1D O(3) non-linear σ-model is a model system for future quantum lattice simulations of other asymptotically-free theories, such as non-Abelian gauge theories. We find that utilizing dimensional reduction can make efficient use of two-dimensional layouts presently available on cold atom quantum simulators. A new definition of the renormalized coupling is introduced, which is applicable to systems with open boundary conditions and can be measured using analog quantum simulators. Monte Carlo and tensor network calculations are performed to determine the quantum resources required to reproduce perturbative short-distance observables. In particular, we show that a rectangular array of 48 Rydberg atoms with existing quantum hardware capabilities should be able to adiabatically prepare low-energy states of the perturbatively-matched theory. Furthermore, these states can then be used to simulate non-perturbative observables in the continuum limit that lie beyond the reach of classical computers.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Enabling Data Exchange and Data Integration with the Common Information Model: An Introduction for Power Systems Engineers and Application Developers

The Common Information Model (CIM) is an open-source information model that is used to model an electrical network and the various equipment used on the network. CIM is widely used for data exchange of bulk transmission power systems and is finding increasing use for distribution systems. Use of a non-proprietary information model (such as CIM) that has been agreed upon and adopted by numerous utilities, vendors, and researchers allows significant reduction in the effort and cost of data integration. Likewise, adoption of open data platforms built around the CIM increases available functionalities for managing and optimizing the smart grid of the future. This report is intended as an introduction to CIM for utility engineers, power systems researchers, and application developers, providing a broad view of the CIM and how particular profiles can be adapted for various use cases. Unlike most other CIM introduction documents and the International Electrotechnical Commission (IEC) standards (which are mostly targeted to an audience of data scientists, enterprise database managers, and platform developers), this report is intended for users of traditional power systems analysis software and other readers without any prior experience with canonical information models, data profiles, or UML modeling.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Extending Parsimonious Bayesian Inference

Parsimonious Bayesian inference is a theoretical framework for efficient data assimilation that seeks to balance increased consistency between predictions and training data against corresponding increases in model complexity. Within this framework, over-training is understood as optimization that encodes excessive information within model parameters while only achieving small improvements between predictions and training data. This project aims to develop practical methods of limiting excess model information during optimization. One key observation is that practical heuristics for parsimonious learning in high-dimensions must balance expressivity, i.e. the ability of the model to capture diverse predictions with only a few non-zero parameters, against discoverability, i.e. the ability to train the model with gradient-based optimization and drive parameters to low information states. As such, we developed logical activation functions that are able to adaptively approximate arbitrary truth tables that define Boolean logic operations within a probabilistic framework. These functions have demonstrated the ability to learn exclusive disjunction (XOR) and conditioned disjunction (if [condition] then [result_if_true] else [result_if_false]) within a single layer of a neural network. To efficiently exploit these activation functions to drive parsimonious learning required several other advances within the domain of variational inference. The most efficient form of complexity suppression is structured sparsification, driving most model parameters to zero while achieving the structural coherence among nonzeros needed for bandwidth reduction. Such models are not only far more efficient at suppressing information-theoretic complexity, they also reduce the other forms of complexity (computations, communication, storage, and the number of dependencies needed to evaluate predictions). Aiming to support enhanced sparsification, this project examined new approaches to high-dimensional variational inference that allow us to calibrate and control parameter uncertainty during optimization. By identifying which parameters can sustain sparsifying perturbations with little impact on prediction quality, we can develop better pruning strategies by framing them as approximate Bayesian inference. These advances also open paths to mitigate concerns with deploying advanced learning methods in resource-constrained environments, such as running models on power-limited or communication-limited devices.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1991

Multi-channel holographic birfurcative neural network system for real-time adaptive EOS data analysis

Viewgraphs on multi-channel holographic bifurcative neural network system for real-time adaptive Earth Observing System (EOS) data analysis are presented. The objective is to research and develop an optical bifurcating neuromorphic pattern recognition system for making optical data array comparisons and to evaluate the use of the system for EOS data classification, reduction, analysis, and other applications.

Liu, Hua-Kuang↗
NASA NTRS2014

Concepts to Support HRP Integration Using Publications and Modeling

Initial efforts are underway to enhance the Human Research Program (HRP)'s identification and support of potential cross-disciplinary scientific collaborations. To increase the emphasis on integration in HRP's science portfolio management, concepts are being explored through the development of a set of tools. These tools are intended to enable modeling, analysis, and visualization of the state of the human system in the spaceflight environment; HRP's current understanding of that state with an indication of uncertainties; and how that state changes due to HRP programmatic progress and design reference mission definitions. In this talk, we will discuss proof-of-concept work performed using a subset of publications captured in the HRP publications database. The publications were tagged in the database with words representing factors influencing health and performance in spaceflight, as well as with words representing the risks HRP research is reducing. Analysis was performed on the publication tag data to identify relationships between factors and between risks. Network representations were then created as one type of visualization of these relationships. This enables future analyses of the structure of the networks based on results from network theory. Such analyses can provide insights into HRP's current human system knowledge state as informed by the publication data. The network structure analyses can also elucidate potential improvements by identifying network connections to establish or strengthen for maximized information flow. The relationships identified in the publication data were subsequently used as inputs to a model captured in the Systems Modeling Language (SysML), which functions as a repository for relationship information to be gleaned from multiple sources. Example network visualization outputs from a simple SysML model were then also created to compare to the visualizations based on the publication data only. We will also discuss ideas for building upon this proof-of-concept work to further support an integrated approach to human spaceflight risk reduction.

Mindock, J.↗
NASA NTRS2015

Risk-Based Causal Modeling of Airborne Loss of Separation

Maintaining safe separation between aircraft remains one of the key aviation challenges as the Next Generation Air Transportation System (NextGen) emerges. The goals of the NextGen are to increase capacity and reduce flight delays to meet the aviation demand growth through the 2025 time frame while maintaining safety and efficiency. The envisioned NextGen is expected to enable high air traffic density, diverse fleet operations in the airspace, and a decrease in separation distance. All of these factors contribute to the potential for Loss of Separation (LOS) between aircraft. LOS is a precursor to a potential mid-air collision (MAC). The NASA Airspace Operations and Safety Program (AOSP) is committed to developing aircraft separation assurance concepts and technologies to mitigate LOS instances, therefore, preventing MAC. This paper focuses on the analysis of causal and contributing factors of LOS accidents and incidents leading to MAC occurrences. Mid-air collisions among large commercial aircraft are rare in the past decade, therefore, the LOS instances in this study are for general aviation using visual flight rules in the years 2000-2010. The study includes the investigation of causal paths leading to LOS, and the development of the Airborne Loss of Separation Analysis Model (ALOSAM) using Bayesian Belief Networks (BBN) to capture the multi-dependent relations of causal factors. The ALOSAM is currently a qualitative model, although further development could lead to a quantitative model. ALOSAM could then be used to perform impact analysis of concepts and technologies in the AOSP portfolio on the reduction of LOS risk.

Geuther, Steven C.↗
NASA NTRS2020

Application of Low-Cost Fine Particulate Mass Monitors to Convert Satellite Aerosol Optical Depth Measurements to Surface Concentrations in North America and Africa

Low-cost particulate mass sensors provide opportunities to assess air quality at unprecedented spatial and temporal resolutions. Established traditional monitoring networks have limited spatial resolution and are simply absent in many major cities across sub-Saharan Africa (SSA). Satellites provide snapshots of regional air pollution but require ground-truthing. Low-cost monitors can supplement and extend data coverage from these sources worldwide, providing a better overall air quality picture. We investigate the utility of such a multi-source data integration approach using two case studies. First, in Pittsburgh, Pennsylvania, both traditional monitoring and dense low-cost sensor networks are compared with satellite aerosol optical depth (AOD) data from NASA's MODIS system, and a linear conversion factor is developed to convert AOD to surface fine particulate matter mass concentration (as PM2.5). With 10 or more ground monitors in Pittsburgh, there is a 2-fold reduction in surface PM2.5 estimation mean absolute error compared to using only a single ground monitor. Second, we assess the ability of combined regional-scale satellite retrievals and local-scale low-cost sensor measurements to improve surface PM2.5 estimation at several urban sites in SSA. In Rwanda, we find that combining local ground monitoring information with satellite data provides a 40 % improvement in surface PM2.5 estimation accuracy with respect to using low-cost ground monitoring data alone. A linear AOD-to-surface-PM2.5 conversion factor developed in Kigali, Rwanda, did not generalize well to other parts of SSA and varied seasonally for the same location, emphasizing the need for ongoing and localized ground-based monitoring, which can be facilitated by low-cost sensors. Overall, we find that combining ground-based low-cost sensor and satellite data, even without including additional meteorological or land use information, can improve and expand spatiotemporal air quality data coverage, especially in data-sparse regions.

AOD↗
NASA NTRS2020

Amino Acid Encoding for Deep Learning Applications

Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-ofthe-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

Hesham ElAbd↗
DOE OSTI2023

Leverage Score Sampling for Parametric PDEs (Final Technical Report)

This final technical report summarizes the accomplishments of work performed under DOE Office of Science Award DE-SC0022266, which is titled “Leverage Score Sampling for Parametric PDEs”. The goal of the project was to extend methods from Randomized Numerical Linear Algebra (RandNLA) to tackle central computational challenges in model order reduction and uncertainty quantification (UQ) for parametric partial differential equations (PDEs). In particular, we sought to use importance sampling methods originally developed for RandNLA to develop sample efficient active learning algorithms for approximating high-dimensional scalar functions, e.g. by polynomials, Gaussian process models, and simple neural networks. Such methods can be immediately applied to developing surrogate models or to approximating quantity of interest (QoI) surfaces. In the context of PDEs, each sample used for learning equates to the solution of the differential equation for a particular set of parameters, so sample efficiency translates to improved computational efficiency for a variety of downstream tasks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Finding Hidden Patterns in High Resolution Wind Flow Model Simulations

Wind flow data is critical in terms of investment decisions and policy making. High resolution data from wind flow model simulations serve as a supplement to the limited resource of original wind flow data collection. Given the large size of data, finding hidden patterns in wind flow model simulations are critical for reducing the dimensionality of the analysis. In this work, we first perform dimension reduction with two autoencoder models: the CNN-based autoencoder (CNN-AE) [1], and hierarchical autoencoder (HIER-AE) [2], and compare their performance with the Principal Component Analysis (PCA). We then investigate the super-resolution of the wind flow data. By training a Generative Adversarial Network (GAN) with 300 epochs, we obtained a trained model with 2× resolution enhancement. We compare the results of GAN with Convolutional Neural Network (CNN), and GAN results show finer structure as expected in the data field images. Also, the kinetic energy spectra comparisons show that GAN outperforms CNN in terms of reproducing the physical properties for high wavenumbers and is critical for analysis where high-wavenumber kinetics play an important role.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Getting lost tracking the carbon footprint of hydropower

In the transition to low-carbon electricity, well-quantified estimates of carbon dynamics are needed to ensure that emissions reduction targets are achieved. Here we review the state of the science on carbon accounting for hydropower reservoirs and identify limitations and future solutions. Nearly all research on reservoir greenhouse-gas (GHG) emissions has focused on individual reservoirs in isolation without considering their position in a freshwater network draining organic matter from upstream watersheds or the coordinated operation of reservoir cascades. Second, carbon inventories have extrapolated from a small, non-probabilistic sample of highly variable measurements of GHG emissions to unsampled reservoirs. A stronger statistical foundation is needed to estimate a global inventory and its uncertainty. Third, attribution to hydropower is based on ranks assigned to reservoir purpose. Instead, the physical influence of hydropower on carbon dynamics could be directly measured. Fourth, current carbon-accounting practices neglect time. A time-varying approach would quantify variation in emissions for electricity portfolios from changes in the fuel mix at different times and account for ancillary services, i.e., the ability to support the grid when variable renewables are not available without using natural gas. Reservoirs also sequester a significant portion of inflowing carbon in sediments and slow the carbon cycle by delaying the return of carbon to the atmosphere for decades to centuries. Together, these refinements would help to illuminate pathways toward meeting energy demand with the longest-possible delay in returning carbon to the atmosphere and without adding ancient sources to the pool of carbon cycling through aquatic ecosystems.

13 HYDRO ENERGY↗