Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Missing data”

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 541 records · Page 30
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Peri-Net-Pro: the neural processes with quantified uncertainty for crack patterns

Abstract This paper develops a deep learning tool based on neural processes (NPs) called the Peri-Net-Pro, to predict the crack patterns in a moving disk and classifies them according to the classification modes with quantified uncertainties. In particular, image classification and regression studies are conducted by means of convolutional neural networks (CNNs) and NPs. First, the amount and quality of the data are enhanced by using peridynamics to theoretically compensate for the problems of the finite element method (FEM) in generating crack pattern images. Second, case studies are conducted with the prototype microelastic brittle (PMB), linear peridynamic solid (LPS), and viscoelastic solid (VES) models obtained by using the peridynamic theory. The case studies are performed to classify the images by using CNNs and determine the suitability of the PMB, LBS, and VES models. Finally, a regression analysis is performed on the crack pattern images with NPs to predict the crack patterns. The regression analysis results confirm that the variance decreases when the number of epochs increases by using the NPs. The training results gradually improve, and the variance ranges decrease to less than 0.035. The main finding of this study is that the NPs enable accurate predictions, even with missing or insufficient training data. The results demonstrate that if the context points are set to the 10th, 100th, 300th, and 784th, the training information is deliberately omitted for the context points of the 10th, 100th, and 300th, and the predictions are different when the context points are significantly lower. However, the comparison of the results of the 100th and 784th context points shows that the predicted results are similar because of the Gaussian processes in the NPs. Therefore, if the NPs are employed for training, the missing information of the training data can be supplemented to predict the results.

Mathematics↗
DOE OSTI2021

Utility-scale Building Type Assignment Using Smart Meter Data

United States building energy use accounted for 40% of total energy use, 74% of peak demand, and $412 billion in 2019. Building energy modeling allows researchers to simulate building physics, gain insights into possible energy/demand saving opportunities, and assess cost-effective resilience amidst climate change. Many building features needed to create building energy models are readily available such as 2D footprints and LiDAR (height). A critical feature that is not generally obtainable is the building type. In partnership with a utility, a years worth of real-world, 15-minute electrical use data has been examined. The smart meter data is compared to 97 different prototype building energy models to assign building type. Real-world considerations including data preparation, quality assurance, and handling of missing values for advanced metering infrastructure data are addressed. Euclidean distance for pattern-matching of energy use, dynamic time warping, and time-window statistics with machine learning are compared for determining building type from measured electricity use.

Bass, Brett↗
DOE OSTI2024

TNet: A Model-Constrained Tikhonov Network Approach for Inverse Problems

Deep learning (DL), in particular deep neural networks, by default is purely data-driven and in general does not require physics. This is the strength of DL but also one of its key limitations when applied to science and engineering problems in which underlying physical properties—such as stability, conservation, and positivity—and accuracy are required. DL methods in their original forms are often not capable of respecting the underlying mathematical models or achieving desired accuracy even in big-data regimes. On the other hand, many data-driven science and engineering problems, such as inverse problems, typically have limited experimental or observational data, and DL would overfit the data in this case. Leveraging information encoded in the underlying mathematical models, we argue, not only compensates for missing information in low data regimes but also provides opportunities to equip DL methods with the underlying physics, hence promoting better generalization. This paper develops a model-constrained DL approach and its variant TNet—a Tikhonov neural network—which are capable of learning not only information hidden in the training data but also in the underlying mathematical models to solve inverse problems governed by partial differential equations in low data regimes. We provide the constructions and some theoretical results for the proposed approaches for both linear and nonlinear inverse problems. Since TNet is designed to learn inverse solutions with Tikhonov regularization, it is interpretable: in fact it recovers Tikhonov solutions for linear cases while potentially approximating Tikhonov solutions for nonlinear inverse problems. We also prove that data randomization can enhance not only the smoothness of the networks but also their generalizations. Comprehensive numerical results confirm the theoretical findings and show that with even as little as 1 training data sample for one-dimensional (1D) deconvolution, 5 for an inverse 2D heat conductivity problem, 100 for inverse initial conditions for a time-dependent 2D Burgers’s equation, and 50 for inverse initial conditions for 2D Navier–Stokes equations, TNet solutions can be as accurate as Tikhonov solutions while being several orders of magnitude faster. Furthermore, this is possible owing to the model-constrained term, replications, and randomization.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Automated defect detection in spent nuclear fuel using combine Cerenkov and Gamma Emission Tomography data

Spent fuel monitoring and characterization has been a pillar of safeguards and nuclear facility monitoring for many years. The Digital Cerenkov Viewing Device (DCVD) has been used since the 1980s as a method of defect detection in spent fuel. In recent years, the accounting of large quantities of spent fuel before storage has renewed interest in this relatively quick and inexpensive method. This has an impact not only in safeguards, but also in for nuclear power facilities, as accounting can be a long and arduous process. Additionally, the DCVD demonstrates limited accuracy in more complex cases such as substitution of a fuel rod with steel, or a partial defect detection. A second method, Gamma Emission Tomography (GET) has been explored as an improved defect detection method, but is much more expensive and invasive than DCVD. This work identifies deficiencies in both methods, and proposes a combination of data gathered from each method to address these deficiencies for improved spent fuel characterization. Initial results are promising, showing 97% detection of a single missing fuel rod when the data types are combined, versus approximately 50% and 70%respectively for DCVD and GET data on their own. Finally, these classification results are obtained with algorithms derived from facial recognition and applied to this problem, yielding unique accuracy in near real-time while also maintaining the information barrier between output and measurement desired in safeguards.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
NASA NTRS2018

History and Status of ALSEP and the Apollo Lunar Data Project

A suite of automated scientific instruments (the Apollo Lunar Surface Experiment Package, or ALSEP) was installed at each of the landing sites of Apollo 12, 14, 15, 16, and 17 from 1969 to 1972. They operated from deployment until decommissioning on 30 September 1977. These data were continuously transmitted to Earth and saved on the Range Tapes, which were recorded at the Manned Space Flight Network stations. These data were also broken out by experiment and sent to the experiment Principal Investigators on what were called the P.I. Tapes. Starting in April 1973 the Range Tape data were stored in digital format on 7-track magnetic tapes, the ARCSAV Tapes. In February 1976, the handling of the Range Tapes was transferred to UT Galveston. They produced 9-track tapes referred to as the Work Tapes. Following the Apollo program the Range and ARCSAV tapes, which were never archived, were lost. The Work Tapes were archived at the National Space Science Data Center (NSSDC). Some investigators archived their individual experiment data with NSSDC as well, but much of the data had minimal documentation, were not in digital form, or were stored in difficult to translate formats. Data from many experiments were never delivered to the NSSDC. The Lunar Data Project was started to address the problem of both missing and not readily usable data. Our effort has resulted in recovery of some of the ARCSAV tapes, recovery and digitization of a large volume of Apollo scientific and technical documentation, and restoration of many ALSEP and other Apollo data collections. Restoration involves deciphering formats, assembling necessary ancillary data (metadata), and packaging data in digital format to be archived with the Planetary Data System (PDS). Recovery of the data from the ARCSAV tapes involved having the tapes read on special equipment and extracting the individual experiment data out of the integrated data stream. We will report on the history and status of the various recovery efforts.

Work Tapes↗
DOE OSTI2021

12 C(e,e'pN) measurements of short range correlations in the tensor-to-scalar interaction transition region

High-momentum configurations of nucleon pairs at short-distance are probed using measurements of the $^{12}$C$(e,e'p)$ and $^{12}$C$(e,e'pN)$ reactions (where $N$ is either $n$ or $p$), at high-$Q^2$ and $x_B>1.1$. The data span a missing-momentum range of 300--1000 MeV/c and are predominantly sensitive to the transition region of the strong nuclear interaction from a Tensor to Scalar interaction. The data are well reproduced by theoretical calculations using the Generalized Contact Formalism with both chiral and phenomenological nucleon-nucleon ($NN$) interaction models. This agreement suggests that the measured high missing-momentum protons up to $1000$ MeV/c predominantly belong to short-ranged correlated (SRC) pairs. The measured $^{12}$C$(e,e'pN)$ / $^{12}$C$(e,e'p)$ and $^{12}$C$(e,e'pp)$ / $^{12}$C$(e,e'pn)$ cross-section ratios are consistent with a decrease in the fraction of proton-neutron SRC pairs and increase in the fraction of proton-proton SRC pairs with increasing missing momentum. This confirms the transition from an isospin-dependent tensor $NN$ interaction at $\sim 400$ MeV/c to an isospin-independent scalar interaction at high-momentum around $\sim 800$ MeV/c as predicted by theoretical calculation.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2020

Updates to Relevance Vector Machine: Multiclass Classification, Variable Selection, and Proof-of-Concept Application to Safeguards Fresh Fuel Verification using List-Mode Neutron Collar Data

To expand the capabilities of safeguards authorities to verify the integrity of fresh fuel assemblies, Oak Ridge National Laboratory has retrofit the existing electronics of the JCC-71 uranium neutron coincidence collar, which contains 18 3 He neutron detectors and an external 241 AmLi(α, n) neutron interrogation source arranged to surround a fresh nuclear fuel assembly. The new electronics system allows analysts to record list-mode neutron multiplicity data in addition to the singles and doubles rates that are currently measured. Based on previous proof-of-concept research, analysis of these new data will identify off-normal fuel configurations in an assembly and characterize or localize the specific partial fuel defects. The purpose of this report it to document the analysis algorithm development and then to demonstrate its capability for the safeguards verification of fresh fuel assemblies using list mode neutron collar data. To analyze the complex list-mode data collected with the upgraded uranium neutron collar, multivariate classification algorithms are being developed using a novel classification method, the relevance vector machine. This approach may be applied to multiclass problems to estimate the probability that test data belongs to one of many possible classes of data. In addition, our method identifies the most useful variables/channels for making predictions, which illuminates the basis for the model’s predictions, and this interpretability is largely unique among data analytics methods. Variable selection occurs during model training and parameter tuning and does not need any external hyperparameter tuning routines. Finally, we apply the modified relevance vector machine to a simulated dataset of list-mode neutron collar data generated with the radiation transport code MCNP. The method can correctly identify off-normal fuel configurations, categorize the data according to four fuel defect scenarios, and rank the channels in the data according to prediction utility. For nuclear safeguards applications, it is concluded that this method has the potential to increase the sensitivity and reliability to detect missing fuel rods from a standard 17 x 17 Pressurized Water Reactor (PWR) fresh fuel assembly. Within this analysis, “off-normal” (i.e., missing fuel rods) were correctly classified in 17 simulated test scenarios with one quarter (25%) of the fresh fuel rods missing using a training data set of 58 simulated measurements.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2015

Review, evaluation, and discussion of the challenges of missing value imputation for mass spectrometry-based label-free global proteomics

In this review, we apply selected imputation strategies to label-free liquid chromatography–mass spectrometry (LC–MS) proteomics datasets to evaluate the accuracy with respect to metrics of variance and classification. We evaluate several commonly used imputation approaches for individual merits and discuss the caveats of each approach with respect to the example LC–MS proteomics data. In general, local similarity-based approaches, such as the regularized expectation maximization and least-squares adaptive algorithms, yield the best overall performances with respect to metrics of accuracy and robustness. However, no single algorithm consistently outperforms the remaining approaches, and in some cases, performing classification without imputation sometimes yielded the most accurate classification. Thus, because of the complex mechanisms of missing data in proteomics, which also vary from peptide to protein, no individual method is a single solution for imputation. In summary, on the basis of the observations in this review, the goal for imputation in the field of computational proteomics should be to develop new approaches that work generically for this data type and new strategies to guide users in the selection of the best imputation for their dataset and analysis objectives.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Online PMU Missing Value Replacement Via Event-Participation Decomposition

We introduce a new method for online Phasor Measurement Unit (PMU) missing value replacement. Our approach allows us to decompose PMU event responses into a non-dynamic component (denoted the participation factor) that can be inferred directly from the past and a dynamic component that can be inferred directly from all other PMUs (denoted the event strength). When missing values occur, we can use these two components, which do not rely on the missing index, to estimate the correct value. The method is extremely fast and can easily be used for online applications. Furthermore, extensive testing on real power system event data reveals that our approach achieves state-of-the-art performance in terms of Mean Absolute Percent Errors (MAPEs) for PMU data dropped during event periods. Here, the method also yields an interpretable and simplified view of events for further analysis and applications. The method relies only on PMU data and does not take outside information such as network topology.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2021

Physics-Informed Neural Networks for Heat Transfer Problems

Abstract Physics-informed neural networks (PINNs) have gained popularity across different engineering fields due to their effectiveness in solving realistic problems with noisy data and often partially missing physics. In PINNs, automatic differentiation is leveraged to evaluate differential operators without discretization errors, and a multitask learning problem is defined in order to simultaneously fit observed data while respecting the underlying governing laws of physics. Here, we present applications of PINNs to various prototype heat transfer problems, targeting in particular realistic conditions not readily tackled with traditional computational methods. To this end, we first consider forced and mixed convection with unknown thermal boundary conditions on the heated surfaces and aim to obtain the temperature and velocity fields everywhere in the domain, including the boundaries, given some sparse temperature measurements. We also consider the prototype Stefan problem for two-phase flow, aiming to infer the moving interface, the velocity and temperature fields everywhere as well as the different conductivities of a solid and a liquid phase, given a few temperature measurements inside the domain. Finally, we present some realistic industrial applications related to power electronics to highlight the practicality of PINNs as well as the effective use of neural networks in solving general heat transfer problems of industrial complexity. Taken together, the results presented herein demonstrate that PINNs not only can solve ill-posed problems, which are beyond the reach of traditional computational methods, but they can also bridge the gap between computational and experimental heat transfer.

Engineering↗
DOE OSTI2018

Network Scale Ubiquitous Volume Estimation Using Tree-Based Ensemble Learning Methods

Currently real-time, ubiquitous volume data for roadway networks remains the key missing dimension in traffic operations. Most volume data are average annual daily traffic (AADT) measures derived from the Highway Performance Monitoring System (HPMS). Although methods to factor the AADT to hourly averages for typical day of week exist, actual volume data is limited to a sparse collection of locations in which volumes are continuously recorded. This paper explores the use of state-of-art machine learning techniques to estimate accurate real-time volume measures that span the highway network providing ubiquitous coverage in space, and point-in-time measures for a specific date and time. Three tree-based ensemble learning models, random forest (RF), gradient boost machine (GBM), and extreme gradient boost (XGBoost), were tested for volume estimation by learning from combined dataset of commercial probe data provided by TomTom, the FHWA's Travel Monitoring Analysis System (TMAS) data, and other infrastructure attributes such as number of lanes, speed limit, and weather. The methods were tested on major corridors and freeways in the metropolitan area of Denver. All three machine learning methods were able to provide hourly volume estimates 24 hours a day, 7 days a week, and 365 days a year with around 18% mean absolute error to true volume and about 5% of error with respect to roadway capacity. The low error measures allow the potential application by transportation agencies.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2020

Network-Scale Ubiquitous Volume Estimation Using Tree-Based Ensemble Learning Methods

Currently ubiquitous volume data for roadway networks remains the key missing dimension in traffic operations. Most volume data are average annual daily traffic (AADT) measures derived from the Highway Performance Monitoring System (HPMS). Although methods to factor the AADT to hourly averages for typical day of week exist, actual volume data is limited to a sparse collection of locations in which volumes are continuously recorded. This paper/poster explores the use of state-of-art machine learning techniques to estimate accurate volume measures that span the highway network providing ubiquitous coverage in space, and point-in-time measures for a specific date and time. Three tree-based ensemble learning models, random forest (RF), gradient boost machine (GBM), and extreme gradient boost (XGBoost), were tested for volume estimation by learning from combined dataset of commercial probe data provided by TomTom, the FHWA's Travel Monitoring Analysis System (TMAS) data, and other infrastructure attributes such as number of lanes, speed limit, and weather. The methods were tested on major corridors and freeways in the metropolitan area of Denver. All three machine learning methods were able to provide hourly volume estimates 24 hours a day, 7 days a week, and 365 days a year with around 18% mean absolute error to true volume and about 5% of error with respect to roadway capacity. The low error measures allow the potential application by transportation agencies.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2014

Software-Based Challenges of Developing the Future Distribution Grid

The software that the utility industry currently uses may be insufficient to analyze the distribution grid as it rapidly modernizes to include active resources such as distributed generation, switch and voltage control, automation, and increasingly complex loads. Although planners and operators have traditionally viewed the distribution grid as a passive load, utilities and consultants increasingly need enhanced analysis that incorporates active distribution grid loads in order to ensure grid reliability. Numerous commercial and open-source tools are available for analyzing distribution grid systems. These tools vary in complexity from providing basic load-flow and capacity analysis under steady-state conditions to time-series analysis and even geographical representations of dynamic and transient events. The need for each type of analysis is not well understood in the industry, nor are the reasons that distribution analysis requires different techniques and tools both from those now available and from those used for transmission analysis. In addition, there is limited understanding of basic capability of the tools and how they should be practically applied to the evolving distribution system. The study reviews the features and state of the art capability of current tools, including usability and visualization, basic analysis functionality, advanced analysis including inverters, and renewable generation and load modeling. We also discuss the need for each type of distribution grid system analysis. In addition to reviewing basic functionality current models, we discuss dynamics and transient simulation in detail and draw conclusions about existing software?s ability to address the needs of the future distribution grid as well as the barriers to modernization of the distribution grid that are posed by the current state of software and model development. Among our conclusions are that accuracy, data transfer, and data processing abilities are key to future distribution grid modeling, and measured data sources are a key missing element . Modeling tools need to be calibrated based on measured grid data to validate their output in varied conditions such as high renewables penetration and rapidly changing topology. In addition, establishing a standardized data modeling format would enable users to transfer data among tools to take advantage of different analysis features. ?

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2013

The MCDB System for Management and Analysis of Petabyte-Scale Uncertain Data

Analysts working with very large data sets often use statistical models to “guess” at unknown, inaccurate, or missing information associated with the data. For example, a distant object viewed through an optical lens will have its position slightly shifted by imperfections in the lens. Thus, rather than considering the object’s observed position to be absolutely correct, it makes sense to take into account the lens’s imperfections to obtain a probabilistic guess as to the object’s true position. For another example, it might be important to associate some sort of error distribution with each of the individual sensors in an array of magnetometers. This error distribution may be complex and include spatially-driven covariances, because errors in nearby sensors are likely to be correlated (caused, for example, by the presence of some nearby, fixed metal object). This project is concerned with the design and implementation of a prototype data management system called the Monte Carlo Database System, or MCDB for short. MCDB allows an expert-level analyst or statistician to attach arbitrary stochastic models to very large data sets in order to “guess” the values for unknown or inaccurate data, such as the actual position of the observed object in the lens example above. When the resulting data set is analyzed, the underlying stochastic models are used to generate hundreds or thousands of possible data set instances, and each of those possible instances is analyzed separately by MCDB. Thus, MCDB does not just give a single answer to the analysis, but it actually gives an empirical distribution of query results that embody the underlying uncertainty, and can in turn be analyzed using standard statistical techniques. The stochastic models in MCDB are implemented as user-defined, external C++ libraries called Variable Generation functions (VG functions for short). Because the VG function interface is exceedingly general, it allows MCDB to be used in a very wide variety of application domains, in conjunction with virtually any stochastic model. Ultimately, we ended up calling the system that we developed over the course of the project SimSQL, to denote the fact that it allowed for stochastic simulation of data. The software developed during the lifetime of the project is available for download from the SimSQL project website: http://cmj4.web.rice.edu/SimSQL/SimSQL.html.

42 ENGINEERING↗
DOE OSTI2026

Merged Aerosol Value-Added Product Report

The Merged Aerosol Value-Added Product (VAP) simplifies scientists’ use of Atmospheric Radiation Measurement (ARM) User Facility aerosol data by performing several tedious, time-consuming tasks for the users. First, the VAP identifies the best data available when multiple datastreams exist for a single geophysical quantity so that ARM users do not have to research this for themselves. Second, the VAP consolidates multiple ARM aerosol datastreams into a single file for ARM data users so that they do not have to download, open, and read multiple files for their analysis. Next, the VAP transforms all measurements onto a common one-hour timestamp. The one-hour resolution matches the time resolution of the slowest instrument. Instruments with faster sampling rates than one measurement per hour are averaged over the time interval. Finally, the VAP reads the QA/QC variables and marks data with known issues as missing, so that users do not have to spend excessive time cleaning data. This includes incorporating Data Quality Reports (DQRs) that exist at the time when the VAP data is generated. DQRs are reports filed by instrument mentors or data users that indicate a problem with the output data of individual instruments.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2010

Regional Distribution of Forest Height and Biomass from Multisensor Data Fusion

Elevation data acquired from radar interferometry at C-band from SRTM are used in data fusion techniques to estimate regional scale forest height and aboveground live biomass (AGLB) over the state of Maine. Two fusion techniques have been developed to perform post-processing and parameter estimations from four data sets: 1 arc sec National Elevation Data (NED), SRTM derived elevation (30 m), Landsat Enhanced Thematic Mapper (ETM) bands (30 m), derived vegetation index (VI) and NLCD2001 land cover map. The first fusion algorithm corrects for missing or erroneous NED data using an iterative interpolation approach and produces distribution of scattering phase centers from SRTM-NED in three dominant forest types of evergreen conifers, deciduous, and mixed stands. The second fusion technique integrates the USDA Forest Service, Forest Inventory and Analysis (FIA) ground-based plot data to develop an algorithm to transform the scattering phase centers into mean forest height and aboveground biomass. Height estimates over evergreen (R2 = 0.86, P < 0.001; RMSE = 1.1 m) and mixed forests (R2 = 0.93, P < 0.001, RMSE = 0.8 m) produced the best results. Estimates over deciduous forests were less accurate because of the winter acquisition of SRTM data and loss of scattering phase center from tree ]surface interaction. We used two methods to estimate AGLB; algorithms based on direct estimation from the scattering phase center produced higher precision (R2 = 0.79, RMSE = 25 Mg/ha) than those estimated from forest height (R2 = 0.25, RMSE = 66 Mg/ha). We discuss sources of uncertainty and implications of the results in the context of mapping regional and continental scale forest biomass distribution.

Yu, Yifan↗