Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “DATA 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 883 records · Page 49

Architectural Implications of Neural Network Inference for High Data-Rate, Low-Latency Scientific Applications

With more scientific fields relying on neural networks (NNs) to process data incoming at extreme throughputs and latencies, it is crucial to develop NNs with all their parameters stored on-chip. In many of these applications, there is not enough time to go off-chip and retrieve weights. Even more so, off-chip memory such as DRAM does not have the bandwidth required to process these NNs as fast as the data is being produced (e.g., every 25 ns). As such, these extreme latency and bandwidth requirements have architectural implications for the hardware intended to run these NNs: 1) all NN parameters must fit on-chip, and 2) codesigning custom/reconfigurable logic is often required to meet these latency and bandwidth constraints. In our work, we show that many scientific NN applications must run fully on chip, in the extreme case requiring a custom chip to meet such stringent constraints.

Weng, Olivia↗

Conjugate-Gradient Neural Networks in Classification of Multisource and Very-High-Dimensional Remote Sensing Data

Application of neural networks to classification of remote sensing data is discussed. Conventional two-layer backpropagation is found to give good results in classification of remote sensing data but is not efficient in training. A more efficient variant, based on conjugate-gradient optimization, is used for classification of multisource remote sensing and geographic data and very-high-dimensional data. The conjugate-gradient neural networks give excellent performance in classification of multisource data, but do not compare as well with statistical methods in classification of very-high-dimentional data.

Benediktsson, J. A.↗

Situational Awareness of Network System Roles (SANSR)

In a large enterprise it is difficult for cyber security analysts to know what services and roles every machine on the network is performing (e.g. file server, domain name server, email server). Understanding the roles of the systems in the network provides analysts with a situational awareness that will allow them to detect consequential changes in the network, initiate an incident response plan, and optimize their security posture. Using the network flow data, already collected by most enterprises, we developed a tool that enables analysts to automatically detect/classify services and roles of every machine that’s operating on a network (e.g. file server, domain name server, email server) for better situational awareness of potential threats to the network. his tool queries Elasticsearch for network flow data, creates a temporal behavior model of each system, uses unsupervised machine learning to cluster the models with a set of labeled temporal behavior models, and the resulting information can be printed to the console or programmatically accessed. The results include the likelihood that a machine has a labeled role and lists other machines that are most similar in behavior.

Huffer, KellyM.↗

Real-time Intrusion Detection for High-bandwidth Research Networks using Unsupervised Deep Learning

Nowadays, data generated by large-scale scientific experiments are on the scale of petabytes per month. These data are transferred through dedicated high-bandwidth networks across distributed collaboration sites for processing, storage, and analysis. Like general purpose networks, research networks experience attacks. However, monitoring traffic in such high-bandwidth network environments is challenging with conventional deep packet inspection (DPI)- based network intrusion detection systems (NIDS). Recently, machine learning (ML) algorithms have shown potential in detecting anomalous patterns with better generalizability and faster speed. Compared to the general purpose networks, traffic movement in research network is less volatile and more likely to benefit from the ML-based anomaly analysis. We present ThunderSecure, an efficient AI-powered system for detecting anomalies in 100G research networks. ThunderSecure implements a high-throughput packet processing infrastructure using multi-cores and GPUs. It takes time-series statistics from a variety of network measurements and uses 2D Convolutional Neural Network (CNN) to explore both spatial and temporal dependencies in network data stream. Patterns of science data traffic are learned with a one-class neural network, which blends an Adversarial Autoencoder (AAE) with Gaussian mixture density estimation. Testing traffic flows exhibiting significant deviations from the learned baseline of normality are marked as anomalies. We trained ThunderSecure on hundreds of billions of science data packets recorded from a 100G research network at Fermi National Accelerator Laboratory. The detection performance was evaluated on traffic captured from the same research network days and weeks after the training with different types of attack flows injected. To the best of our knowledge, this is the first ML work in the area of network anomaly detection that has been validated on such extreme scale datasets. Results show that ThunderSecure can recognize science data traffic that are captured long after the training and make nearly certain detection on those with anomalous flows injected, even in the case when the anomaly-to-normal mixing ratio is 0.1%.

Gong, Qian↗

Processed Lab Data for Neural Network-Based Shear Stress Level Prediction

Machine learning can be used to predict fault properties such as shear stress, friction, and time to failure using continuous records of fault zone acoustic emissions. The files are extracted features and labels from lab data (experiment p4679). The features are extracted with a non-overlapping window from the original acoustic data. The first column is the time of the window. The second and third columns are the mean and the variance of the acoustic data in this window, respectively. The 4th-11th column is the the power spectrum density ranging from low to high frequency. And the last column is the corresponding label (shear stress level). The name of the file means which driving velocity the sequence is generated from. Data were generated from laboratory friction experiments conducted with a biaxial shear apparatus. Experiments were conducted in the double direct shear configuration in which two fault zones are sheared between three rigid forcing blocks. Our samples consisted of two 5-mm-thick layers of simulated fault gouge with a nominal contact area of 10 by 10 cm^2. Gouge material consisted of soda-lime glass beads with initial particle size between 105 and 149 micrometers. Prior to shearing, we impose a constant fault normal stress of 2 MPa using a servo-controlled load-feedback mechanism and allow the sample to compact. Once the sample has reached a constant layer thickness, the central block is driven down at constant rate of 10 micrometers per second. In tandem, we collect an AE signal continuously at 4 MHz from a piezoceramic sensor embedded in a steel forcing block about 22 mm from the gouge layer The data from this experiment can be used with the deep learning algorithm to train it for future fault property prediction.

15 GEOTHERMAL ENERGY↗

Regional Heatwave Prediction Using Graph Neural Network and Weather Station Data

Abstract Heatwaves lead to catastrophic consequences on public health and the economy. Accurate and timely predictions of regional heatwaves can improve climate preparedness and foster decision‐making to alleviate the burdens due to climate change. In this paper, we propose a heatwave prediction algorithm based on a novel deep learning model, that is, Graph Neural Network (GNN). This new GNN framework can provide real time warnings of the sudden occurrence of regional heatwaves with high accuracy at lower costs of computation and data collection. In addition, its interpretable structure unravels the spatiotemporal patterns of regional heatwaves and helps to enrich our understanding of the general climate dynamics and the causal influences between locations. The proposed GNN framework can be applied for the detection and prediction of other extreme or compound climate events, which calls for future studies.

54 ENVIRONMENTAL SCIENCES↗

Data Fusion via Neural Network Entropy Minimization for Target Detection and Multi-Sensor Event Classification

Broadly applicable solutions to multimodal and multisensory fusion problems across domains remain a challenge because effective solutions often require substantive domain knowledge and engineering. The chief questions that arise for data fusion are in when to share information from different data sources, and how to accomplish the integration of information. The solutions explored in this work remain agnostic to input representation and terminal decision fusion approaches by sharing information through the learning objective as a compound objective function. The objective function this work uses assumes a one-to-one learning paradigm within a one-to-many domain which allows the assumption that consistency can be enforced across the one-to-many dimension. The domains and tasks we explore in this work include multi-sensor fusion for seismic event location and multimodal hyperspectral target discrimination. We find that our domain- informed consistency objectives are challenging to implement in stable and successful learning because of intersections between inherent data complexity and practical parameter optimization. While multimodal hyperspectral target discrimination was not enhanced across a range of different experiments by the fusion strategies put forward in this work, seismic event location benefited substantially, but only for label-limited scenarios.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A comparison of neural network architectures for data-driven reduced-order modeling

The popularity of deep convolutional autoencoders (CAEs) has engendered new and effective reduced-order models (ROMs) for the simulation of large-scale dynamical systems. Despite this, it is still unknown whether deep CAEs provide superior performance over established linear techniques or other network-based methods in all modeling scenarios. To elucidate this, the effect of autoencoder architecture on its associated ROM is studied through the comparison of deep CAEs against two alternatives: a simple fully connected autoencoder, and a novel graph convolutional autoencoder. Through benchmark experiments, it is shown that the superior autoencoder architecture for a given ROM application is highly dependent on the size of the latent space and the structure of the snapshot data, with the proposed architecture demonstrating benefits on data with irregular connectivity when the latent space is sufficiently large.

42 ENGINEERING↗

Transparent application-layer/os deeper packet inspector

A computer-implemented method of deep packet inspection (DPI) in a network is provided. The method comprises collecting data packets comprising a number of traffic flows from a number of devices via a number of traffic taps and classifying each traffic flow according to data about network protocol layers of the packets comprising the traffic flow. Application layer metadata is extracted from the packets. Traffic flow classification data and the extracted metadata are ingested into a data cluster and normalized. The normalized classification data and extracted metadata is then correlated to other data sets.

Urias, Vincent↗

Fast BPM data distribution for global orbit feedback using commercial gigabit ethernet technology

In order to correct beam perturbations in RHIC around 10Hz, a new fast data distribution network was required to deliver BPM position data at rates several orders of magnitude above the capability of the existing system. The urgency of the project limited the amount of custom hardware that could be developed, which dictated the use of as much commercially available equipment as possible. The selected architecture uses a custom hardware interface to the existing RHIC BPM electronics together with commercially available Gigabit Ethernet switches to distribute position data to devices located around the collider ring. Using the minimum Ethernet packet size and a field programmable gate array (FPGA) based state machine logic instead of a software based driver, real-time and deterministic data delivery is possible using Ethernet. The method of adapting this protocol for low latency data delivery, bench testing of Ethernet hardware, and the logic to construct Ethernet packets using FPGA hardware will be discussed. A robust communications system using almost all commercial off-the-shelf equipment was developed in under a year which enabled retrofitting of the existing RHIC BPM system to provide 10 KHz data delivery for a global orbit feedback scheme using 72 BPMs. Total latencies from data acquisition at the BPMs to delivery at the controller modules, including very long transmission distances, were kept under 100 {micro}s, which provide very little phase error in correcting the 10 Hz oscillations. Leveraging off of the speed of Gigabit Ethernet and wide availability of Ethernet products enabled this solution to be fully implemented in a much shorter time and at lower cost than if a similar network was developed using a proprietary method.

43 PARTICLE ACCELERATORS↗

The Medical Science DMZ

Objective We describe use cases and an institutional reference architecture for maintaining high-capacity, data-intensive network flows (e.g., 10, 40, 100 Gbpsþ) in a scientific, medical context while still adhering to security and privacy laws and regulations. Materials and Methods High-end networking, packet filter firewalls, network intrusion detection systems. Results We describe a “Medical Science DMZ” concept as an option for secure, high-volume transport of large, sensitive data sets between research institutions over national research networks. Discussion The exponentially increasing amounts of “omics” data, the rapid increase of high-quality imaging, and other rapidly growing clinical data sets have resulted in the rise of biomedical research “big data.” The storage, analysis, and network resources required to process these data and integrate them into patient diagnoses and treatments have grown to scales that strain the capabilities of academic health centers. Some data are not generated locally and cannot be sustained locally, and shared data repositories such as those provided by the National Library of Medicine, the National Cancer Institute, and international partners such as the European Bioinformatics Institute are rapidly growing. The ability to store and compute using these data must therefore be addressed by a combination of local, national, and industry resources that exchange large data sets. Maintaining data-intensive flows that comply with HIPAA and other regulations presents a new challenge for biomedical research. Recognizing this, we describe a strategy that marries performance and security by borrowing from and redefining the concept of a “Science DMZ”—a framework that is used in physical sciences and engineering research to manage high-capacity data flows. Conclusion By implementing a Medical Science DMZ architecture, biomedical researchers can leverage the scale provided by high-performance computer and cloud storage facilities and national high-speed research networks while preserving privacy and meeting regulatory requirements.

97 MATHEMATICS AND COMPUTING↗

Transfer Learning of High-Fidelity Opacity Spectra in Autoencoders and Surrogate Models

Simulations of high energy density physics are expensive, largely in part for the need to produce nonlocal thermodynamic equilibrium opacities. High-fidelity spectra may reveal new physics in the simulations not seen with low-fidelity spectra, but the cost of these simulations also scales with the level of fidelity of the opacities being used. Neural networks are capable of reproducing these spectra, but neural networks need data to train them, which limits the level of fidelity of the training data. Here this article demonstrates that it is possible to reproduce high-fidelity spectra with median errors in the realm of 3%–4% using as few as 50 samples of high-fidelity Krypton data by performing transfer learning on a neural network trained on many times more low-fidelity data.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

SPAN: Ocean science

The Space Physics Analysis Network (SPAN) is a multi-mission, correlative data comparison network which links space and Earth science research and data analysis computers. It provides a common working environment for sharing computer resources, sharing computer peripherals, solving proprietary problems, and providing the potential for significant time and cost savings for correlative data analysis. This is one of a series of discipline-specific SPAN documents which are intended to complement the SPAN primer and SPAN Management documents. Their purpose is to provide the discipline scientists with a comprehensive set of documents to assist in the use of SPAN for discipline specific scientific research.

Thomas, Valerie L.↗

Aftershock Identification Using a Paired Neural Network Applied to Constructed Data

This report is intended to detail the findings of our investigation of the applicability of machine learning to the task of aftershock identification. The ability to automatically identify nuisance aftershock events to reduce analyst workload when searching for events of interest is an important step in improving nuclear monitoring capabilities and while waveform cross - correlation methods have proven successful, they have limitations (e.g., difficulties with spike artifacts, multiple aftershocks in the same window) that machine learning may be able to overcome. Here we apply a Paired Neural Network (PNN) to a dataset consisting of real, high quality signals added to real seismic noises in order to work with controlled, labeled data and establish a baseline of the PNN's capability to identify aftershocks. We compare to waveform cross - correlation and find that the PNN performs well, outperforming waveform cross - correlation when classifying similar waveform pairs, i.e., aftershocks.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

A Data Gathering Scheme in Wireless Sensor Networks Based on Synchronization of Chaotic Spiking Oscillator Networks

This paper studies chaos-based data gathering scheme in multiple sink wireless sensor networks. In the proposed scheme, each wireless sensor node has a simple chaotic oscillator. The oscillators generate spike signals with chaotic interspike intervals, and are impulsively coupled by the signals via wireless communication. Each wireless sensor node transmits and receives sensor information only in the timing of the couplings. The proposed scheme can exhibit various chaos synchronous phenomena and their breakdown phenomena, and can effectively gather sensor information with the significantly small number of transmissions and receptions compared with the conventional scheme. Also, the proposed scheme can flexibly adapt various wireless sensor networks not only with a single sink node but also with multiple sink nodes. This paper introduces our previous works. Through simulation experiments, we show effectiveness of the proposed scheme and discuss its development potential.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Displaying Composite and Archived Soundings in the Advanced Weather Interactive Processing System

In a previous task, the Applied Meteorology Unit (AMU) developed spatial and temporal climatologies of lightning occurrence based on eight atmospheric flow regimes. The AMU created climatological, or composite, soundings of wind speed and direction, temperature, and dew point temperature at four rawinsonde observation stations at Jacksonville, Tampa, Miami, and Cape Canaveral Air Force Station, for each of the eight flow regimes. The composite soundings were delivered to the National Weather Service (NWS) Melbourne (MLB) office for display using the National version of the Skew-T Hodograph analysis and Research Program (NSHARP) software program. The NWS MLB requested the AMU make the composite soundings available for display in the Advanced Weather Interactive Processing System (AWIPS), so they could be overlaid on current observed soundings. This will allow the forecasters to compare the current state of the atmosphere with climatology. This presentation describes how the AMU converted the composite soundings from NSHARP Archive format to Network Common Data Form (NetCDF) format, so that the soundings could be displayed in AWl PS. The NetCDF is a set of data formats, programming interfaces, and software libraries used to read and write scientific data files. In AWIPS, each meteorological data type, such as soundings or surface observations, has a unique NetCDF format. Each format is described by a NetCDF template file. Although NetCDF files are in binary format, they can be converted to a text format called network Common data form Description Language (CDL). A software utility called ncgen is used to create a NetCDF file from a CDL file, while the ncdump utility is used to create a CDL file from a NetCDF file. An AWIPS receives soundings in Binary Universal Form for the Representation of Meteorological data (BUFR) format (http://dss.ucar.edu/docs/formats/bufr/), and then decodes them into NetCDF format. Only two sounding files are generated in AWIPS per day. One file contains all of the soundings received worldwide between 0000 UTC and 1200 UTC, and the other includes all soundings between 1200 UTC and 0000 UTC. In order to add the composite soundings into AWIPS, a procedure was created to configure, or localize, AWIPS. This involved modifying and creating several configuration text files. A unique fourcharacter site identifier was created for each of the 32 soundings so each could be viewed separately. The first three characters were based on the site identifier of the observed sounding, while the last character was based on the flow regime. While researching the localization process for soundings, the AMU discovered a method of archiving soundings so old soundings would not get purged automatically by AWl PS. This method could provide an alternative way of localizing AWl PS for composite soundings. In addition, this would allow forecasters to use archived soundings in AWIPS for case studies. A test sounding file in NetCDF format was written in order to verify the correct format for soundings in AWIPS. After the file was viewed successfully in AWIPS, the AMU wrote a software program in the Tool Command Language/Tool Kit (Tcl/Tk) language to convert the 32 composite soundings from NSHARP Archive to CDL format. The ncgen utility was then used to convert the CDL file to a NetCDF file. The NetCDF file could then be read and displayed in AWIPS.

Barrett, Joe H., III↗

Model-parallel Fourier neural operators as learned surrogates for large-scale parametric PDEs

Fourier neural operators (FNOs) are a recently introduced neural network architecture for learning solution operators of partial differential equations (PDEs), which have been shown to perform significantly better than comparable deep learning approaches. Once trained, FNOs can achieve speed-ups of multiple orders of magnitude over conventional numerical PDE solvers. However, due to the high dimensionality of their input data and network weights, FNOs have so far only been applied to two-dimensional or small three-dimensional problems. To remove this limited problem-size barrier, we propose a model-parallel version of FNOs based on domain-decomposition of both the input data and network weights. Here, we demonstrate that our model-parallel FNO is able to predict time-varying PDE solutions of over 2.6 billion variables on Perlmutter using up to 512 A100 GPUs and show an example of training a distributed FNO on the Azure cloud for simulating multiphase CO 2 dynamics in the Earth’s subsurface.

58 GEOSCIENCES↗

Solving the Orszag–Tang vortex magnetohydrodynamics problem with physics-constrained convolutional neural networks

We study the 2D Orszag–Tang vortex magnetohydrodynamics (MHD) problem through the use of physics-constrained convolutional neural networks (PCNNs) for forecasting the density, ρ, and the magnetic field, B, as well as the prediction of B given the velocity field v of the fluid. In addition to translation equivariance from the convolutional architecture, other physics constraints were embedded: absence of magnetic monopoles, non-negativity of ρ, use of only relevant variables, and the periodic boundary conditions of the problem. The use of only relevant variables and the hard constraint of non-negative ρ were found to facilitate learning greatly. The divergenceless condition ∇·B=0 was implemented as a hard constraint up to machine precision through the use of a magnetic potential to define B=∇×A. Residual networks and data augmentation were also used to improve performance. This allowed for some of the residual models to function as surrogate models and provide reasonably accurate simulations. For the prediction task, the PCNNs were evaluated against a physics-informed neural network, which had the ideal MHD induction equation as a soft constraint. Several models were able to generate highly accurate fields, which are visually almost indistinguishable and have low mean squared error. Only methods with built-in hard constraints produced physical fields with ∇·B=0. The use of PCNNs for MHD has the potential to produce physically consistent real-time simulations to serve as virtual diagnostics in cases where inferences must be made with limited observables.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗