Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network regularization”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Unified Medical Language System resources improve sieve-based generation and Bidirectional Encoder Representations from Transformers (BERT)–based ranking for concept normalization

Concept normalization, the task of linking phrases in text to concepts in an ontology, is useful for many downstream tasks including relation extraction, information retrieval, etc. We present a generate-and-rank concept normalization system based on our participation in the 2019 National NLP Clinical Challenges Shared Task Track 3 Concept Normalization. The shared task provided 13 609 concept mentions drawn from 100 discharge summaries. We first design a sieve-based system that uses Lucene indices over the training data, Unified Medical Language System (UMLS) preferred terms, and UMLS synonyms to generate a list of possible concepts for each mention. We then design a listwise classifier based on the BERT (Bidirectional Encoder Representations from Transformers) neural network to rank the candidate concepts, integrating UMLS semantic types through a regularizer. Our generate-and-rank system was third of 33 in the competition, outperforming the candidate generator alone (81.66% vs 79.44%) and the previous state of the art (76.35%). During postevaluation, the model’s accuracy was increased to 83.56% via improvements to how training data are generated from UMLS and incorporation of our UMLS semantic type regularizer. Analysis of the model shows that prioritizing UMLS preferred terms yields better performance, that the UMLS semantic type regularizer results in qualitatively better concept predictions, and that the model performs well even on concepts not seen during training. Our generate-and-rank framework for UMLS concept normalization integrates key UMLS features like preferred terms and semantic types with a neural network–based ranking model to accurately link phrases in text to UMLS concepts.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2025

An adaptive and stability-promoting layerwise training approach for sparse deep neural network architecture

This work presents a two-stage adaptive framework for progressively developing deep neural network (DNN) architectures that generalize well for a given training data set. In the first stage, a layerwise training approach is adopted where a new layer is added each time and trained independently by freezing parameters in the previous layers. We impose desirable structures on the DNN by employing manifold regularization, sparsity regularization, and physics-informed terms. We introduce a ε – δ – stability-promoting concept as a desirable property for a learning algorithm and show that employing manifold regularization yields a ε – δ stability-promoting algorithm. Further, we also derive the necessary conditions for the trainability of a newly added layer and investigate the training saturation problem. In the second stage of the algorithm (post-processing), a sequence of shallow networks is employed to extract information from the residual produced in the first stage, thereby improving the prediction accuracy. Numerical investigations on prototype regression and classification problems demonstrate that the proposed approach can outperform fully connected DNNs of the same size. Moreover, by equipping the physics-informed neural network (PINN) with the proposed adaptive architecture strategy to solve partial differential equations, we numerically show that adaptive PINNs not only are superior to standard PINNs but also produce interpretable hidden layers with provable stability. As a result, we also apply our architecture design strategy to solve inverse problems governed by elliptic partial differential equations.

42 ENGINEERING↗
NASA NTRS2016

The Plant Phenology Monitoring Design for the National Ecological Observatory Network

Phenology is an integrative science that comprises the study of recurring biological activities or events. In an era of rapidly changing climate, the relationship between the timing of those events and environmental cues such as temperature, snowmelt, water availability, or day length are of particular interest. This article provides an overview of the observer-based plant phenology sampling conducted by the U.S. National Ecological Observatory Network (NEON), the resulting data, and the rationale behind the design. Trained technicians will conduct regular in situ observations of plant phenology at all terrestrial NEON sites for the 30-yr life of the observatory. Standardized and coordinated data across the network of sites can be used to quantify the direction and magnitude of the relationships between phenology and environmental forcings, as well as the degree to which these relationships vary among sites, among species, among phenophases, and through time. Vegetation at NEON sites will also be monitored with tower-based cameras, satellite remote sensing, and annual high-resolution airborne remote sensing. Ground-based measurements can be used to calibrate and improve satellite-derived phenometrics. NEON's phenology monitoring design is complementary to existing phenology research efforts and citizen science initiatives throughout the world and will produce interoperable data. By collocating plant phenology observations with a suite of additional meteorological, biophysical, and ecological measurements (e.g., climate, carbon flux, plant productivity, population dynamics of consumers) at 47 terrestrial sites, the NEON design will enable continental-scale inference about the status, trends, causes, and ecological consequences of phenological change.

phenology↗
NASA NTRS1998

Results of the NY-Alesund ozone measurements intercomparison NAOMI

Worldwide, about ten Differential Absorption Lidars are used for long-term monitoring of stratospheric ozone. These systems are an important component of the Network for the Detection of Stratospheric Change. Although DIALs are self-calibrating in principle, regular intercomparisons with other ozone-lidars, microwave radiometers or ozone-sondes are highly desirable to ensure high data quality at a well known level. The Network for the Detection of Stratospheric Change (NDSC) validation policy suggests that such intercomparisons be "blind", meaning all participants submit their data to an impartial referee, without seeing results from the other participants. Here we report on the "blind" intercomparison taking place from January 20th to February 10th 1998 at Ny-Alesund, Spitsbergen (78.92 deg N, 11.95 deg E). Participating groups were from the Alfred Wegener Institute, Potsdam, operating the NDSC DIAL system at Ny-Alesund, from the University of Bremen operating the NDSC microwave radiometer for ozone profiling at Ny-Alesund, and the NASA Goddard Space Flight Center group with the "NDSC travelling standard" STROZ-LITE. The first author acted as the impartial referee. Also used for the intercomparison were data from ECC-6A/Vaisala RS80 ozone sondes routinely launched at Ny-Alesund by the AWI group. A 1% KI solution (3 ml) and the 1986 ECC pump correction (1.092 at 5 hPa) are used. The ECC-data were available to all participants during the campaign and thus were not "blind". Table 1 summarizes the expected performance of the instruments participating in the ozone intercomparison reported in this paper.

Steinbrecht, W.↗
NASA NTRS2009

Wireless Avionics Packet to Support Fault Tolerance for Flight Applications

In this protocol and packet format, data traffic is monitored by all network interfaces to determine the health of transmitter and subsystems. When failures are detected, the network inter face applies its recover y policies to provide continued service despite the presence of faults. The protocol, packet format, and inter face are independent of the data link technology used. The current demonstration system supports both commercial off-the-shelf wireless connections and wired Ethernet connections. Other technologies such as 1553 or serial data links can be used for the network backbone. The Wireless Avionics packet is divided into three parts: a header, a data payload, and a checksum. The header has the following components: magic number, version, quality of service, time to live, sending transceiver, function code, payload length, source Application Data Interface (ADI) address, destination ADI address, sending node address, target node address, and a sequence number. The magic number is used to identify WAV packets, and allows the packet format to be updated in the future. The quality of service field allows routing decisions to be made based on this value and can be used to route critical management data over a dedicated channel. The time to live value is used to discard misrouted packets while the source transceiver is updated at each hop. This information is used to monitor the health of each transceiver in the network. To identify the packet type, the function code is used. Besides having a regular data packet, the system supports diagnostic packets for fault detection and isolation. The payload length specifies the number of data bytes in the payload, and this supports variable-length packets in the network. The source ADI is the address of the originating interface. This can be used by the destination application to identify the originating source of the packet where the address consists of a subnet, subsystem class within the subnet, a subsystem unit, and the local ADI number. The destination ADI is used to route the packet to its ultimate destination. At each hop, the sending interface uses the destination address to determine the next node for the data. The sending node is the node address of the interface that is broadcasting the packet. This field is used to determine the health of the subsystem that is sending the packet. In the case of a packet that traverses several intermediate nodes, it may be the node address of the intermediate node. The target node is the node address of the next hop for the packet. It may be an intermediate node, or the final destination for the packet. The sequence number is used to identify duplicate packets. Because each interface has multiple transceivers, the same packet will appear at both receivers. The sequence number allows the interface to correlate the reception and forward a single, unique packet for additional processing. The subnet field allows data traffic to be partitioned into segregated local networks to support large networks while keeping each subnet at a manageable size. This also keeps the routing table small enough so routing can be done by a simple table lookup in an FPGA device. The subsystem class identifies members of a set of redundant subsystems, and, in a hot standby configuration, all members of the subsystem class will receive the data packets. Only the active subsystem will generate data traffic. Specific units in a class of redundant units can be identified and, if the hot standby configuration is not used, packets will be directed to a specific subsystem unit.

Block, Gary L.↗
NASA NTRSDate not supplied

Delay Tolerant, Radio Frequency Identification (RFID )-enabled Sensing

Radio Frequency Identification (RFID) technology offers a completely passive method to transmit fixed data sequences from an RFID tag, which typically doesn’t have its own power supply, to an interrogator. Radio frequency (RF) energy harvested from the interrogator is rectified by the tag and used to charge an integrated circuit (IC). The IC then modulates the received signal with the data stored on the tag and reflects the energy back to the interrogator. RFID has seen great proliferation in terrestrial inventory management applications, and it has recently made the jump to spaceflight applications onboard the International Space Station, augmenting an existing optical bar‐code infrastructure for tracking supplies. A number of advanced automated logistics management (ALM) concepts employing RFID are currently being developed and evaluated, including so‐called “smart” shelves, cubbies, and trash receptacles using low‐power, embeddable RFID interrogators. An infrastructure where both crew members and robotic assistants, such as autonomous free flyers, are similarly equipped with small RFID interrogators seems likely. It therefore behooves us to consider extending this infrastructure beyond ALM to applications such as low power, embedded sensing. Typically, data on an RFID tag can only be written by an RFID interrogator, using interrogator energy. In recent years, however, a few efforts have focused on using that energy to drive data acquisition from the tag IC, allowing the tag to modify its stored data sequence with sensor data before replying to an interrogator. In this way, the tag can act as a completely passive sensing device. One problem exists with this approach, however: the tag cannot gather data when an interrogator is not present. Thus, strictly passive RFID sensing tags cannot gather data at regular intervals, in the manner of a typical wireless sensor network, without careful, and impractical, planning of mobile interrogator movements. To address this shortcoming, we look to a recent advance in RFID technology which allows an external microprocessor to power the tag IC and write directly into its RFID memory using a wired serial interface. In this paradigm, data gathering is driven by a small, on‐board power supply (using batteries or harvested energy), and data transfer is provided passively through the RFID interrogation service. Since communication typically consumes the lion’s share of power in WSNs, such a technique has the potential to enable extremely long‐lived, embedded wireless sensing when used with extremely low‐current microcontrollers. Since the communication channel is only open when an interrogator is present and actively interrogating the RFID sensing tag, transport of periodically‐sampled sensor data presents itself as a delay/disruption‐tolerant networking (DTN) problem. In this paper, we present the design of a DTN‐like overlay on the common EPC Global, Class 1, Generation 2 RFID standard. This overlay allows seamless, guaranteed data transfer using the EPC Global protocol, supporting extremely low‐power, embedded sensing using an infrastructure likely to be already in place for ALM applications. We evaluate a prototype implementation of a complete end‐to‐end system using a robotic RFID interrogation agent, and we present future directions for the development of this sensing technique.

Raymond S. Wagner↗
NASA NTRSDate not supplied

Deep Space navigation for the BioSentinel spacecraft science orbit

BioSentinel is an astrobiology small spacecraft mission. The payload consists of two parts, the first has optical and microfluidics sensors, and the second is a Linear Energy Transfer spectrometer that has the objective to measure deep space radiation from events such as coronal mass ejections. The goal of the mission is to observe potential DNA damage due to the radiation in heliocentric space on the living organism Saccharomyces cerevisiae, which is a budding yeast. Two types of this living organism are included in the payload. The first is a natural type that is more radiation tolerant, while the second is a mutant strain that has a deficiency in a gene that allows DNA repair once damage occurs. The impact caused by the radiation on the DNA is compared to an identical sample aboard the International Space Station, as well as another identical sample at a laboratory on the ground. The BioSentinel mission consists of a 6U CubeSat currently ,as of January 2024, active in heliocentric orbit. The spacecraft was launched aboard the first SLS flight as part of the Artemis-I campaign in November 2022. After successful deployment from the launch vehicle, it performed a lunar flyby with an altitude of 406 km. The delta-V imparted by the flyby provided the necessary energy to achieve a heliocentric orbit, in an Earth-trailing pattern. The navigation analysis consisted of a Kalman-filter that utilized data from the Deep Space Network and the ESA Estrack network. All those antennas were needed since the Artemis-1 campaign included the deployment of several other cubesats, therefore the scheduling process required more antenna assets than usual due to simultaneous demands from various missions. The processed tracking data was later also refined with a smoother in order to obtain a more accurate solution. The type of tracking data included TCP, Sequential Range, Doppler and Range formats. The solar radiation pressure coefficient, as well as the delta-V from the deployment and the flyby were modeled to obtain suitable solutions that could decrease the position and velocity uncertainties at several steps along the mission concept of operations. The final product each time resulted in updated ephemeris files that were used by the mission and the antenna networks as the mission progressed. Once in the final science orbit, the utilized antennas are only from the DSN network and the data format is bounded to just TCP. Regular orbit determination is performed, every two weeks. The spacecraft is in a nominal well-known orbit, performing regular operations. This paper includes an analysis of the final science orbit, the techniques and procedures utilized to perform orbit determination and a description of the overall navigation campaign produced during the mission and, more specifically, during the final science operations in Deep Space.

BioSentinel↗
NASA NTRS2023

Performance of the IVS R1 and R4 Sessions

The International VLBI Service for Geodesy and Astrometry (IVS) has observed the weekly IVS-R1 (R1) and IVS-R4 (R4) series of sessions since 2002. These regular series are generally stable in the sense that the networks were designed to be similar from week to week. The uniformity of these series has allowed researchers to conduct many scientific investigations. The IVS also observes with other networks that allow continued sampling of data from all VLBI stations, but the R1 and R4 observing sessions are a dominant part of overall VLBI observing accounting for 1841 sessions out of a total of 3129 24-hour sessions from 2002.0 to 2020.0 (where the last R1 and R4 in 2019 was on December 30 and December 26, respectively). In this paper, we investigate the evolution of these series in terms of their observing networks. We also discuss the construction of the R1 and R4 networks and the scheduling of these sessions. The performance of these networks in terms of the formal precision of polar motion have improved by factors of 2–3 over the period from 2002.0 to 2018.0. UT1 precision improved by a factor of about 1.2–1.5. The main reason for this improvement is the increased size of the networks. We also discuss the effect on this improvement arising from changes in the data rate and the number of observed sources. There is some degradation in performance after 2018 that is most likely due to a decline in the number of available network stations.

Cynthia C. Thomas↗
DOE OSTI2024

Surrogate construction via weight parameterization of residual neural networks

Surrogate model development is a critical step for uncertainty quantification or other sample-intensive tasks for complex computational models. Here, in this work, we develop a multi-output surrogate form using a class of neural networks (NNs) that employ shortcut connections, namely Residual NNs (ResNets). ResNets are known to regularize the surrogate learning problem and improve the efficiency and accuracy of the resulting surrogate. Inspired by the continuous, Neural ODE analogy, we augment ResNets with weight parameterization strategy with respect to ResNet depth. Weight-parameterized ResNets regularize the NN surrogate learning problem and allow better generalization with a drastically reduced number of learnable parameters. We demonstrate that weight-parameterized ResNets are more accurate and efficient than conventional feed-forward multi-layer perceptron networks. We also compare various options for parameterization of the weights as functions of ResNet depth. We demonstrate the results on both synthetic examples and a large scale earth system model of interest.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Enhancing Streamflow Forecast and Extracting Insights Using Long-Short Term Memory Networks With Data Integration at Continental Scales

Recent observations with varied schedules and types (moving average, snapshot, or regularly spaced) can help to improve streamflow forecasts, but it is challenging to integrate them effectively. Based on a long short-term memory (LSTM) streamflow model, we tested multiple versions of a flexible procedure we call data integration (DI) to leverage recent discharge measurements to improve forecasts. DI accepts lagged inputs either directly or through a convolutional neural network unit. DI ubiquitously elevated streamflow forecast performance to unseen levels, reaching a record continental-scale median Nash-Sutcliffe Efficiency coefficient value of 0.86. Integrating moving-average discharge, discharge from the last few days, or even average discharge from the previous calendar month could all improve daily forecasts. Directly using lagged observations as inputs was comparable in performance to using the convolutional neural network unit. Importantly, we obtained valuable insights regarding hydrologic processes impacting LSTM and DI performance. Before applying DI, the base LSTM model worked well in mountainous or snow-dominated regions, but less well in regions with low discharge volumes (due to either low precipitation or high precipitation-energy synchronicity) and large interannual storage variability. DI was most beneficial in regions with high flow autocorrelation: it greatly reduced baseflow bias in groundwater-dominated western basins and also improved peak prediction for basins with dynamical surface water storage, such as the Prairie Potholes or Great Lakes regions. However, even DI cannot elevate performance in high-aridity basins with 1-day flash peaks. Despite this limitation, there is much promise for a deep-learning-based forecast paradigm due to its performance, automation, efficiency, and flexibility.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Learning the simplicity of scattering amplitudes

The simplification and reorganization of complex expressions lies at the core of scientific progress, particularly in theoretical high-energy physics. This work explores the application of machine learning to a particular facet of this challenge: the task of simplifying scattering amplitudes expressed in terms of spinor-helicity variables. We demonstrate that an encoder-decoder transformer architecture achieves impressive simplification capabilities for expressions composed of handfuls of terms. Lengthier expressions are implemented in an additional embedding network, trained using contrastive learning, which isolates subexpressions that are more likely to simplify. The resulting framework is capable of reducing expressions with hundreds of terms—a regular occurrence in quantum field theory calculations—to vastly simpler equivalent expressions. Starting from lengthy input expressions, our networks can generate the Parke-Taylor formula for five-point gluon scattering, as well as new compact expressions for five-point amplitudes involving scalars and gravitons.

Cheung, Clifford [California Institute of Technolo↗
NASA NTRS1993

On the accuracy of ERS-1 orbit predictions

Since the launch of ERS-1, the D-PAF (German Processing and Archiving Facility) provides regularly orbit predictions for the worldwide SLR (Satellite Laser Ranging) tracking network. The weekly distributed orbital elements are so called tuned IRV's and tuned SAO-elements. The tuning procedure, designed to improve the accuracy of the recovery of the orbit at the stations, is discussed based on numerical results. This shows that tuning of elements is essential for ERS-1 with the currently applied tracking procedures. The orbital elements are updated by daily distributed time bias functions. The generation of the time bias function is explained. Problems and numerical results are presented. The time bias function increases the prediction accuracy considerably. Finally, the quality assessment of ERS-1 orbit predictions is described. The accuracy is compiled for about 250 days since launch. The average accuracy lies in the range of 50-100 ms and has considerably improved.

Koenig, Rolf↗
NASA NTRS1988

Terminal attractors for addressable memory in neural networks

A new type of attractors - terminal attractors - for an addressable memory in neural networks operating in continuous time is introduced. These attractors represent singular solutions of the dynamical system. They intersect (or envelope) the families of regular solutions while each regular solution approaches the terminal attractor in a finite time period. It is shown that terminal attractors can be incorporated into neural networks such that any desired set of these attractors with prescribed basins is provided by an appropriate selection of the weight matrix.

Zak, Michail↗
DOE OSTI2025

Qubit Regularization of Quantum Field Theories

To study quantum field theories on a quantum computer, we must begin with Hamiltonians defined on a finite-dimensional Hilbert space and then take appropriate limits. This approach can be seen as a new type of regularization for quantum field theories, which we refer to as qubit regularization. A related finite-dimensional regularization, known as the D-theory approach, was proposed long ago as a general framework for all quantum field theories. In this framework, the dimensionality of the local Hilbert space at each spatial point can increase as needed through an additional flavor index. To reproduce asymptotically free QFTs, most studies assume that qubit-regularized theories require extending the local Hilbert space to infinity. However, contrary to this common belief, recent discoveries in (1+1) dimensions have revealed two examples where asymptotic freedom appears to emerge within a strictly finite-dimensional local Hilbert space through a novel renormalization group (RG) flow. These findings motivate further investigation into whether asymptotically free gauge theories could also emerge within a strictly finite-dimensional local Hilbert space. To support these explorations, we propose an orthonormal basis called the monomer-dimer-tensor-network (MDTN) basis and use it to construct new types of qubit-regularized lattice gauge theories.

Chandrasekharan, Shailesh [Duke Univ., Durham, NC ↗
DOE OSTI2022

Power quality disturbances diagnosis: A 2D densely connected convolutional network framework

The fast and accurate diagnosis of power quality disturbances (PQD) aids in avoiding shutdowns and unnecessary procedures, concerning electric energy distribution systems. As such, a number of techniques have been tested and applied in order to reach this objective. Majority of the techniques applied are two-step based. On the first step, power quality disturbances features are extracted. Second step, considering features extracted, disturbance classification is implemented. Recently, relevant literature has presented data-driven signal processing-based approaches, as deep convolutional neural networks (DCNN), which can implement both processing steps while providing automated recognition of patterns and outliers in data. However, not considered by state-of-art, power quality disturbances are evolving in nature, while all possible regularities might not be represented in the dataset. In this work a 2 Dimension Densely Connected Convolutional Network (2D-DenseNet) framework is presented. Further, a case study with synthetic disturbance events are analyzed. Easy-to-implement formulation, built on the 2D-DenseNet, without hard-to-design parameters, highlight potential aspects for real-life implementation.

42 ENGINEERING↗
DOE OSTI2022

Inverse Analysis with Variational Autoencoders: A Comparison of Shallow and Deep Networks

Inverse problems are applied to determine unknown properties by matching observational data with a physical model that often takes many parameters as input. To overcome the underconstrained nature of inverse problems and achieve good performance, an approach is presented involving regularization with a technique known as a variational autoencoder (VAE), which is trained to map a high-dimensional parameter space with a complex structure to a low-dimensional latent space with a simple structure. We apply this approach to unconditioned realizations of the parameters (heterogeneous hydraulic fields) for a hydrogeological inverse problem. Two types of hydraulic conductivity fields are used to evaluate the characterization for the different levels of heterogeneity complexity of the physical inputs. This approach keeps the computational cost of generating the training data low. The reason is unconditioned realizations neither rely on the observational data used to perform the inverse analysis nor require any groundwater flow model forward runs. In addition, this approach applies regularization on a low-dimensional latent space from the VAE and increases optimization efficiency through automatic differentiation. Furthermore, two different neural network (NN) structures are tested for their utility in using the VAE for inverse analysis. The performance of a deep, convolutional neural network strongly depends on the dimensionality of the latent space, which requires tuning. In contrast, a shallow, dense neural network provides consistently accurate characterization without tuning. Furthermore, our approach evaluates the advantages of the shallow, dense neural network over the deep, convolutional one and enables future application to a wide range of inverse problems.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2000

NASA Goddard Space Flight Center

The NASA SLR Operational Center is responsible for: 1) NASA SLR network control, sustaining engineering, and logistics; 2) ILRS mission operations; and 3) ILRS and NASA SLR data operations. NASA SLR network control and sustaining engineering tasks include technical support, daily system performance monitoring, system scheduling, operator training, station status reporting, system relocation, logistics and support of the ILRS Networks and Engineering Working Group. These activities ensure the NASA SLR systems are meeting ILRS and NASA mission support requirements. ILRS mission operations tasks include mission planning, mission analysis, mission coordination, development of mission support plans, and support of the ILRS Missions Working Group. These activities ensure than new mission and campaign requirements are coordinated with the ILRS. Global Normal Points (NP) data, NASA SLR FullRate (FR) data, and satellite predictions are managed as part of data operations. Part of this operation includes supporting the ILRS Data Formats and Procedures Working Group. Global NP data operations consist of receipt, format and data integrity verification, archiving and merging. This activity culminates in the daily electronic transmission of NP files to the CDDIS. Currently of all these functions are automated. However, to ensure the timely and accurate flow of data, regular monitoring and maintenance of the operational software systems, computer systems and computer networking are performed. Tracking statistics between the stations and the data centers are compared periodically to eliminate lost data. Future activities in this area include sub-daily (i.e., hourly) NP data management, more stringent data integrity tests, and automatic station notification of format and data integrity issues.

Carter, David↗