Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Edge computing”

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 739 records · Page 41

Hyperparameter Optimization and Feature Inclusion in Graph Neural Networks for Spiking Implementation

Graph convolutional networks leverage both graph structures and features on nodes and edges for improved learning performance in comparison with classical machine learning approaches. Spiking neuromorphic computers natively implement network-like computation and have been shown to be successful at implementing graph learning without features. Incorporating graph features brings the challenge of efficient feature representation and balancing the contribution of topology and features in learning. In this work, we present our design of a simulated network of spiking neurons to perform semi-supervised learning on graph data using both the graph structure and the node features. We explore various design choices, present preliminary results, and discuss the opportunities for using neuromorphic computers for this task in the future.

Cong, Guojing↗

Orbit-product representation and correction of Gaussian belief propagation

We present a new interpretation of Gaussian belief propagation (GaBP) based on the 'zeta function' representation of the determinant as a product over orbits of a graph. We show that GaBP captures back-tracking orbits of the graph and consider how to correct this estimate by accounting for non-backtracking orbits. We show that the product over non-backtracking orbits may be interpreted as the determinant of the non-backtracking adjacency matrix of the graph with edge weights based on the solution of GaBP. An efficient method is proposed to compute a truncated correction factor including all non-backtracking orbits up to a specified length.

99 GENERAL AND MISCELLANEOUS↗

Advanced High Torque Density Magnetically Geared Generator

The objective of this research project was to investigate the fundamental torque density limitations of multistage flux focusing magnetic gears and magnetically geared direct-drive generators for wind turbines. Magnetic gears create non-contact speed amplification by modulating the air-gap magnetic field harmonics created by magnetic rotors. By integrating a stator into the magnetic gear structure a very high torque density direct-drive generator can be created. Magnetic geared systems offer unique capabilities that are currently not available using existing mechanically geared systems, such as overload protection. Magnetic gears also offer the potential for higher reliability, higher efficiency and quieter operation than is currently achievable using a mechanically geared generator. By increasing reliability and design life magnetic gears could decrease the overall life-cycle costs when compared with equivalent mechanical or direct-drive drivetrains. This research project investigated and experimentally testing two sub-scale magnetic gearboxes (0.12 m diameter) and two large diameter magnetic gearboxes (0.56m and 0.63m diameter). The goal of the project was to demonstrate a torque density >300Nm/L for a multistage magnetic gearbox with integrated stator. A nested multistage magnetic gear structure was first proposed and studied. However, it was shown that the nested design greatly increased mechanical complexity and therefore a two-stage series typology was utilized. A multi-stage magnetic gear with a 59:1 gear ratio was successfully experimentally tested. A peak torque of 4.25 kN·m was measured and this resulted in an overall active region torque density of 141 N·m/L (when including support rings). The torque density was significantly reduced from what was computed using 2-D electromagnetic analysis. This reduction was due to the 3-D edge effects, increased power losses and the inclusion of additional mechanical support rings as well as other mechanical assembly design changes. At large diameters contiguous lamination parts could not be fabricated and so a segmented laminated rotor design had to be used. This reduced the mechanical strength of the design when compared to small diameter magnetic gears and resulted in the need for the addition of mechanical support rings to prevent deflection. The final experimentally tested design also had a high torque ripple and the power loss within the magnetic gear did not change significantly with load. This therefore resulted in the efficiency being high only at peak loads. Despite not experimentally achieving the 2-D calculated high torque density the analysis has shown that the problems encountered with the scaling can be overcome with careful design changes. For example, understanding the unique mechanical deflection requirements during the electromagnetic design of the large diameter magnetic gears is key to achieving both a high torque density based on the electromagnetic analysis and a high torque density based on the mechanical support needs. Furthermore, ensuring that end plates are made of non-conductive material and minimizing other conductive support elements within the magnetic structure is a key requirement for reducing eddy current losses. In summary, if the mechanical design can be made more robust and eddy current loss mitigation strategies are used in the mechanical design phase then the magnetic gear is capable of being scaled up whilst operating at a very high torque density. This will then enable it to be potentially performance and cost competitive with alternative direct-drive and mechanically geared wind turbine drivetrains.

17 WIND ENERGY↗

Electric Distribution System Planning with DERs - High-level Assessment of Tools and Methods

This document addresses analytical tools and methods to support electric distribution system planning with distributed energy resources (DERs) and grid modernization, drawing on leading-edge research and examples from around the United States. In this report, tools include computer models and software, as well as other analytical aids and practices that support distribution system planning. At its core, distribution system planning is about supporting investment decisions. Enhanced and integrated distribution system planning allows utilities to move from a passive reactive role to smart proactive planning with respect to DERs. Distribution system planning can increase transparency around utility distribution system investments before they are brought to the regulatory body for cost recovery, can support deferring traditional infrastructure investments, and can provide information and incentives to customers and third parties about where DERs can help the grid. A growing number of states are beginning to consider a comprehensive distribution system planning process that addresses the costs and benefits of DERs, reflects more choices coming from the customer/third-party domains, is integrated with resource and transmission planning, and incorporates planning scenarios that reflect the dynamic and changing nature of the electricity system

24 POWER TRANSMISSION AND DISTRIBUTION↗

Cathodic Protection Modeling for Hanford Underground Double-Shell Tank Farms

Hanford stores millions of gallons of radioactive and chemically hazardous waste from the production of weapon materials in tank farms consisting of underground carbon-steel storage tanks surrounded by reinforced concrete. Six of these Hanford tank farms use double-shell storage tanks (DSTs). The DST farms were constructed from 1968 to 1986 with a planned 40–50 year design life, so some are already operating beyond their initial life expectancy. Ultrasonic testing (UT) has indicated significant thinning on the bottom of the secondary (outer) liner of these tanks, believed to arise from groundwater intrusion driving concrete side corrosion. There is no direct access to the steel/concrete interface between the tank and the concrete pad, making it difficult to apply a chemical-based mitigation strategy or to conduct repairs, but cathodic protection (CP) is a possible method to inhibit further concrete-side corrosion. Hanford already uses CP to protect below grade steel piping within the tank farms and connected to the tanks, but this system was not designed to protect the tank bottoms. CP design must account for the structures surrounding the DSTs, including the steel reinforcing bars (rebar) within the concrete pad and vault, various process lines, and the existing CP system. In this study, finite element analysis (FEA) modeling was carried out to simulate CP protection of 1) a single tank and CP anode to develop options for modeling the rebar and to compare to a simpler circuit model and 2) the entire Hanford AN tank farm as a representative example consisting of seven tanks, associated piping, and both existing and new CP anodes. Both circuit and FEA models predict that significant protective current could be delivered to the bottoms of the tanks with the addition of tank-protection anodes below the depth of the tanks. Simulations with only the existing pipe-protection anodes active confirmed that only a very small current to the tank bottoms is predicted under present conditions. Multiple simplified representations of the dome and wall rebar were tested to reduce the computational complexity of the tank-farm simulations, resulting in modeling the rebar as edge elements with a prescribed effective circumference that matches the real rebar surface area. The geometry of the rebar is also simplified into horizontal hoops around the tank walls and radial rebar over the dome with increased effective circumference to retain the target surface area. This simplification was found to greatly reduce the complexity and solution time of the models without large changes in current distributions, especially to the tank bottom. A range of values were tested for model parameters such as soil and concrete resistivities and polarization resistance to investigate their impact on the current and electric potential distributions. Depending on the parameters used, FEA simulations predict some risk of overprotection, particularly on the piping system; since overprotection can also lead to surface damage associated with hydrogen gas generation at the interface (e.g. hydrogen embrittlement or damage to coatings), this needs to be considered when refining the design of the new CP system. Comparison between the FEA models and the circuit model representation demonstrated that the circuit model could not match the predicted FEA current distribution, even when using the exact same surface areas. This discrepancy appeared to be at least partly attributable to the impact of the relative positions of the tank components and anodes to each other and to the ground surface. The FEA model accounts for the relative positions since it solves the governing equations in three dimensions, but the circuit model cannot account for the positioning. In particular, the circuit model underpredicts the current to the tank bottom and overpredicts the current to the dome compared to FEA for the baseline geometry. The FEA models omitted the electrically isolated rebar in the bottom concrete slab. However, a circuit based stray current model estimated that only 2.1% of the total current through the slab would stray into the rebar, corresponding to ~0.21 A for a target current density of 2 mA/ft2 to the tank bottom. The estimated corrosion driven by this amount of stray current is predicted to yield a lifetime of >400 years for the minimum rebar diameter, assuming an acceptable cross-section area loss of 10%.

d'Entremont, Anna [Savannah River National Laborat↗

Magnet optical and beam matching issues in SNS MEBT

A Medium-Energy Beam Transport (MEBT) line is employed in the SNS linac. The MEBT lattice consists of fourteen electromagnetic quadrupoles and other devices. The quads have very small aspect ratios, and they are densely packed. Significant fringe fields and magnetic interference cause difficulties in beam matching. We have performed 3D simulations of the magnets, computed their optical properties, and compared their performance with what predicted by simple hard edge models. This paper reports our findings and a general solution to the problem.

43 PARTICLE ACCELERATORS↗

Roofline Analysis in the Intel® Advisor to Deliver Optimized Performance for applications on Intel® Xeon Phi™ Processor

In this session we show, in two case studies, how the roofline feature of Intel Advisor has been utilized to optimize the performance of kernels of the XGC1 and PICSAR codes in preparation for Intel Knights Landing architecture. The impact of the implemented optimizations and the benefits of using the automatic roofline feature of Intel Advisor to study performance of large applications will be presented. This demonstrates an effective optimization strategy that has enabled these science applications to achieve up to 4.6 times speed-up and prepare for future exascale architectures. # Goal/Relevance of Session The roofline model [1,2] is a powerful tool for analyzing the performance of applications with respect to the theoretical peak achievable on a given computer architecture. It allows one to graphically represent the performance of an application in terms of operational intensity, i.e. the ratio of flops performed and bytes moved from memory in order to guide optimization efforts. Given the scale and complexity of modern science applications, it can often be a tedious task for the user to perform the analysis on the level of functions or loops to identify where performance gains can be made. With new Intel tools, it is now possible to automate this task, as well as base the estimates of peak performance on measurements rather than vendor specifications. The goal of this session is to demonstrate how the roofline feature of Intel Advisor can be used to balance memory vs. computation related optimization efforts and effectively identify performance bottlenecks. A series of typical optimization techniques: cache blocking, structure refactoring, data alignment, and vectorization illustrated by the kernel cases will be addressed. # Description of the codes ## XGC1 The XGC1 code [3] is a magnetic fusion Particle-In-Cell code that uses an unstructured mesh for its Poisson solver that allows it to accurately resolve the edge plasma of a magnetic fusion device. After recent optimizations to its collision kernel [4], most of the computing time is spent in the electron push (pushe) kernel, where these optimization efforts have been focused. The kernel code scaled well with MPI+OpenMP but had almost no automatic compiler vectorization, in part due to indirect memory addresses and in part due to low trip counts of low-level loops that would be candidates for vectorization. Particle blocking and sorting have been implemented to increase trip counts of low-level loops and improve memory locality, and OpenMP directives have been added to vectorize compute-intensive loops that were identified by Advisor. The optimizations have improved the performance of the pushe kernel 2x on Haswell processors and 1.7x on KNL. The KNL node-for-node performance has been brought to within 30% of a NERSC Cori phase I Haswell node and we expect to bridge this gap by reducing the memory footprint of compute intensive routines to improve cache reuse. ## PICSAR is a Fortran/Python high-performance Particle-In-Cell library targeting at MIC architectures first designed to be coupled with the PIC code WARP for the simulation of laser-matter interaction and particle accelerators. PICSAR also contains a FORTRAN stand-alone kernel for performance studies and benchmarks. A MPI domain decomposition is used between NUMA domains and a tile decomposition (cache-blocking) handled by OpenMP has been added for shared-memory parallelism and better cache management. The so-called current deposition and field gathering steps that compose the PIC time loop constitute major hotspots that have been rewritten to enable more efficient vectorization. Particle communications between tiles and MPI domain has been merged and parallelized. All considered, these improvements provide speedups of 3.1 for order 1 and 4.6 for order 3 interpolation shape factors on KNL configured in SNC4 quadrant flat mode. Performance is similar between a node of cori phase 1 and KNL at order 1 and better on KNL by a factor 1.6 at order 3 with the considered test case (homogeneous thermal plasma).

97 MATHEMATICS AND COMPUTING↗

Anomaly Detection in Dynamic Networks

Anomaly detection in dynamic communication networks has many important security applications. These networks can be extremely large and so detecting any changes in their structure can be computationally challenging; hence, computationally fast, parallelisable methods for monitoring the network are paramount. For this reason the methods presented here use independent node and edge based models to detect locally anomalous substructures within communication networks. As a first stage, the aim is to detect changes in the data streams arising from node or edge communications. Throughout the thesis simple, conjugate Bayesian models for counting processes are used to model these data streams. A second stage of analysis can then be performed on a much reduced subset of the network comprising nodes and edges which have been identified as potentially anomalous in the first stage. The first method assumes communications in a network arise from an inhomogeneous Poisson process with piecewise constant intensity. Anomaly detection is then treated as a changepoint problem on the intensities. The changepoint model is extended to incorporate seasonal behavior inherent in communication networks. This seasonal behavior is also viewed as a changepoint problem acting on a piecewise constant Poisson process. In a static time frame, inference is made on this extended model via a Gibbs sampling strategy. In a sequential time frame, where the data arrive as a stream, a novel, fast Sequential Monte Carlo (SMC) algorithm is introduced to sample from the sequence of posterior distributions of the change points over time. A second method is considered for monitoring communications in a large scale computer network. The usage patterns in these types of networks are very bursty in nature and don’t fit a Poisson process model. For tractable inference, discrete time models are considered, where the data are aggregated into discrete time periods and probability models are fitted to the communication counts. In a sequential analysis, anomalous behavior is then identified from outlying behavior with respect to the fitted predictive probability models. Seasonality is again incorporated into the model and is treated as a changepoint model on the transition probabilities of a discrete time Markov process. Second stage analytics are then developed which combine anomalous edges to identify anomalous substructures in the network.

97 MATHEMATICS AND COMPUTING↗

Multimodal correlative imaging and modelling of phosphorus uptake from soil by hyphae of mycorrhizal fungi

Phosphorus (P) is essential for plant growth. Arbuscular mycorrhizal fungi (AMF) aid its uptake by acquiring P from sources distant from roots in return for carbon. Little is known about how AMF colonise soil pore-space, and models of AMF-enhanced P-uptake are poorly validated. We used synchrotron X-ray computed tomography to visualize mycorrhizas in soil and synchrotron X-ray fluorescence/X-ray absorption near edge structure (XRF/XANES) elemental mapping for P, sulphur (S) and aluminium (Al) in combination with modelling. We found that AMF inoculation had a suppressive effect on colonisation by other soil fungi and identified differences in structure and growth rate between hyphae of AMF and nonmycorrhizal fungi. Our results showed that AMF co-locate with areas of high P and low Al, and preferentially associate with organic-type P species over Al-rich inorganic P. We discovered that AMF avoid Al-rich areas as a source of P. Sulphur-rich regions were found to be correlated with higher hyphal density and an increased organic-associated P-pool, whilst oxidized S-species were found close to AMF hyphae. Increased S oxidation close to AMF suggested the observed changes were microbiome-related. Our experimentally-validated model led to an estimate of P-uptake by AMF hyphae that is an order of magnitude lower than rates previously estimated – a result with significant implications for the modelling of plant–soil–AMF interactions.

79 ASTRONOMY AND ASTROPHYSICS↗

Probing Terra Incognita of Ni–P Catalysts: Operando Explorations during Hydrogen Evolution Reaction

We have developed two Ni phosphide preparation methods allowing operando XAS surface-sensitive studies of well-defined bulk systems. For Ni K-edge XAS, a Ni 2 P phase-pure powder was sintered into a high-density pellet and polished for grazing incidence XAS. Ni sites were mildly affected by the acidic electrolyte prior to the HER, while the applied cathodic potential caused the reduction of Ni surface states beyond the states of as-prepared Ni 2 P. The computed fully H-covered Ni 2 P [0001] model describes the difference in the operando Ni K-edge GIXAS spectrum well. Upon turning the applied bias off, the Ni sites became immediately oxidized, forming NiO on the surface. Thus, the active phase during the HER is covalent Ni 0 close to that in the intermetallic phosphides, and Ni 2+ oxides formed after, and not during, the HER. For P K-edge XAS, Ni foam was phosphorized to form a thin Ni 3 P layer while preserving its high surface area. Upon immersion in the acidic electrolyte, the P sites underwent removal of P 5+ phosphates and formed new P coordination, possibly due to the adsorption of protons from the electrolyte. These new P surface states were not affected by turning the cathodic current on and off as soon as the sample was immersed in the acidic electrolyte. However, the removal of the sample from the electrochemical cell and drying in air resulted in substantial depletion and oxidation of surface P. Echoing the observed Ni site chemistry during HER, XAS and XPS suggest that the in situ active P sites are different from the oxidized P states observed under ex situ conditions.

Kong, Seongyoung [Iowa State Univ., Ames, IA (Unit↗

Band Edge Control of Quasi–2D Metal Halide Perovskites for Blue Light–Emitting Diodes with Enhanced Performance

Perovskite light–emitting diodes (PeLEDs) have received great attention for their potential as next–generation display technology. While remarkable progress has been achieved in green, red, and near–infrared PeLEDs with external quantum efficiencies (EQEs) exceeding 20%, obtaining high performance blue PeLEDs remains a challenge. Poor charge balance due to large charge injection barriers in blue PeLEDs has been identified as one of the major roadblocks to achieve high efficiency. Here band edge control of perovskite emitting layers for blue PeLEDs with enhanced charge balance and device performance is reported. By using organic spacer cations with different dipole moments, that is, phenethyl ammonium (PEA), methoxy phenethyl ammonium (MePEA), and 4–fluoro phenethyl ammonium (4FPEA), the band edges of quasi–2D perovskites are tuned without affecting their band gaps. Furthermore, detailed characterization and computational studies have confirmed the effect of dipole moment modification to be mostly electrostatic, resulting in changes in the ionization energies of ≈0.45 eV for MePEA and ≈ –0.65 eV for 4FPEA based thin films relative to PEA–based thin films. With improved charge balance, blue PeLEDs based on MePEA quasi–2D perovskites show twofold increase of the EQE as compared to the control PEA based devices.

36 MATERIALS SCIENCE↗

Risks of secondary malignancies with heterotopic bone radiation therapy for patients younger than 40 years

Heterotopic ossification (HO) of the bone is defined as a benign condition in which abnormal bone formation occurs in soft tissue. One of the most common prophylactic treatments for HO is radiation therapy (RT). This study retrospectively reviewed 20 patients younger than the age of 40 who received radiation to prevent HO in a single fraction of 7 Gray. The purpose of this study is to assess the risk of a second malignancy in these patients by recreating their treatment fields and contouring organs at risk to estimate the radiation dose absorbed by normal tissues outside the radiation treatment field. Diagnostic computed tomography (CT) scans for each patient were used to recreate treatment fields and to calculate dose to structures of interest. The distance from the field edge to each structure and its depth was recorded. Dose measurements in a water phantom were performed for the range of depths, distances, and field sizes used in the actual treatment plans. Computer-generated doses were compared to estimates based on measurement. The structure dose recorded was the higher dose generated between the 2 methods. Scatter dose was recorded to the rectum, bladder, sigmoid colon, small bowel, ovaries and utero-cervix in female patients, and prostate and gonads in male patients. In some patients, there is considerable dose received by certain organs from scatter because of their proximity to the radiation field. The average dose to the ovarian region was 4.125 Gy with a range of 1.085 to 6.228 Gy. The risk estimate for these patients ranged from 0.16% to 0.93%. The average total lifetime risk estimate for the bladder in all patients is 0.22% and the average total lifetime risk estimate for the remainder organs in all patients is 1.25%. In conclusions, proper shielding created from multileaf collimators (MLCs), blocks, and shields should always be used when possible.

61 RADIATION PROTECTION AND DOSIMETRY↗

Full-wave feasibility study of anti-radar diagnostic of magnetic field based on O-X mode conversion and oblique reflectometry imaging

An innovative millimeter wave diagnostic is proposed to measure the local magnetic field and edge current as a function of the minor radius in the tokamak pedestal region. The idea is to identify the direction of minimum reflectivity at the O-mode cutoff layer. Correspondingly, the transmissivity due to O-X mode conversion is maximum. That direction, and the angular map of reflectivity around it, contains information on the magnetic field vector B at the cutoff layer. Probing the plasma with different wave frequencies provides the radial profile of B. Full-wave finite-element simulations are presented here in 2D slab geometry. Modeling confirms the existence of a minimum in reflectivity that depends on the magnetic field at the cutoff, as expected from mode conversion physics, giving confidence in the feasibility of the diagnostic. The proposed reflectometric approach is expected to yield superior signal-to-noise ratio and to access wider ranges of density and magnetic field, compared with related radiometric techniques that require the plasma to emit electron Bernstein waves. Due to computational limitations, frequencies of 10-20 GHz were considered in this initial study. Frequencies above the edge electron-cyclotron frequency (f > 28 GHz here) would be preferable for the experiment, because the upper hybrid resonance and right cutoff would lie in the plasma, and would help separate the O-mode of interest from spurious X-waves.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

IT roadmap requirements for research and development

In 2009, members representing the programmatic community of the Laboratory presented a request to the Information Architecture Steering Committee (IASC) to form a committee within the Information Architecture Program (IAP) to review and consider the needs of researchers. In response, a Research and Development Subteam under the Infrastructure Team was created. Members included volunteer representatives of groups performing research and development throughout the Laboratory. This paper represents the product of this review. The R&D computing environment has languished while Los Alamos has been very compliance driven for the past few years under a set of directives that have eliminated our ability to innovate and to push the envelope of R&D computing. The appeal for outside collaborations is rapidly becoming tarnished, as colleagues from outside the NNSA environment view our restrictions as increasingly oppressive. We are losing the ability to conduct cutting-edge research and losing money to competitors who are able to quickly react and pursue program dollars while we are still creating compliant but less than useful computing solutions. Research programs depend on collaboration, employees, foreign nationals, and frequent travel, including travel to foreign countries. Solutions need to facilitate these activities and be based on the risk that the activities actually pose to the infrastructure and data at LANL. Furthermore, when problems are solved, it is essential that the solutions are evaluated for cost/benefit to the programs, that programmatic impact is well identified, and that solutions have sufficient long term resource for collaborations.

99 GENERAL AND MISCELLANEOUS↗

Towards quasi-transverse momentum dependent PDFs computable on the lattice

Transverse momentum dependent parton distributions (TMDPDFs) which appear in factorized cross sections involve infinite Wilson lines with edges on or close to the light-cone. Since these TMDPDFs are not directly calculable with a Euclidean path integral in lattice QCD, we study the construction of quasi-TMDPDFs with finite-length spacelike Wilson lines that are amenable to such calculations. We define an infrared consistency test to determine which quasi-TMDPDF definitions are related to the TMDPDF, by carrying out a one-loop study of infrared logarithms of transverse position b T ~Λ QCD -1 , which must agree between them. This agreement is a necessary condition for the two quantities to be related by perturbative matching. TMDPDFs necessarily involve combining a hadron matrix element, which nominally depends on a single light-cone direction, with soft matrix elements that necessarily depend on two light-cone directions. We show at one loop that the simplest definitions of the quasi hadron matrix element, the quasi soft matrix element, and the resulting quasi-TMDPDF all fail the infrared consistency test. Ratios of impact parameter quasi-TMDPDFs still provide nontrivial information about the TMD-PDFs, and are more robust since the soft matrix elements cancel. We show at one loop that such quasi ratios can be matched to ratios of the corresponding TMDPDFs. We also introduce a modified “bent” quasi soft matrix element which yields a quasi-TMDPDF that passes the consistency test with the TMDPDF at one loop, and discuss potential issues at higher orders.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗