Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “classification”

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 577 records · Page 32
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Deep Learning Based Superconducting Radio-Frequency Cavity Fault Classification at Jefferson Laboratory

This work investigates the efficacy of deep learning (DL) for classifying C100 superconducting radio-frequency (SRF) cavity faults in the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab. CEBAF is a large, high-power continuous wave recirculating linac that utilizes 418 SRF cavities to accelerate electrons up to 12 GeV. Recent upgrades to CEBAF include installation of 11 new cryomodules (88 cavities) equipped with a low-level RF system that records RF time-series data from each cavity at the onset of an RF failure. Typically, subject matter experts (SME) analyze this data to determine the fault type and identify the cavity of origin. This information is subsequently utilized to identify failure trends and to implement corrective measures on the offending cavity. Manual inspection of large-scale, time-series data, generated by frequent system failures is tedious and time consuming, and thereby motivates the use of machine learning (ML) to automate the task. This study extends work on a previously developed system based on traditional ML methods (Tennant and Carpenter and Powers and Shabalina Solopova and Vidyaratne and Iftekharuddin, Phys. Rev. Accel. Beams, 2020, 23, 114601), and investigates the effectiveness of deep learning approaches. The transition to a DL model is driven by the goal of developing a system with sufficiently fast inference that it could be used to predict a fault event and take actionable information before the onset (on the order of a few hundred milliseconds). Because features are learned, rather than explicitly computed, DL offers a potential advantage over traditional ML. Specifically, two seminal DL architecture types are explored: deep recurrent neural networks (RNN) and deep convolutional neural networks (CNN). We provide a detailed analysis on the performance of individual models using an RF waveform dataset built from past operational runs of CEBAF. In particular, the performance of RNN models incorporating long short-term memory (LSTM) are analyzed along with the CNN performance. Furthermore, comparing these DL models with a state-of-the-art fault ML model shows that DL architectures obtain similar performance for cavity identification, do not perform quite as well for fault classification, but provide an advantage in inference speed.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Classification Analysis of Southwest Pacific Tropical Cyclone Intensity Changes Prior to Landfall

This study evaluates the ability of a random forest classifier to identify tropical cyclone (TC) intensification or weakening prior to landfall over the western region of the Southwest Pacific Ocean (SWPO) basin. For both Australia mainland and SWPO island cases, when a TC first crosses land after spending ≥24 h over the ocean, the closest hour prior to the intersection is considered as the landfall hour. If the maximum wind speed (V max ) at the landfall hour increased or remained the same from the 24-h mark prior to landfall, the TC is labeled as intensifying and if the V max at the landfall hour decreases, the TC is labeled as weakening. Geophysical and aerosol variables closest to the 24 h before landfall hour were collected for each sample. The random forest model with leave-one-out cross validation and the random oversampling example technique was identified as the best-performing classifier for both mainland and island cases. The model identified longitude, initial intensity, and sea skin temperature as the most important variables for the mainland and island landfall classification decisions. Incorrectly classified cases from the test data were analyzed by sorting the cases by their initial intensity hour, landfall hour, monthly distribution, and 24-h intensity changes. TC intensity changes near land strongly impact coastal preparations such as wind damage and flood damage mitigations; hence, this study will contribute to improve identifying and prioritizing prediction of important variables contributing to TC intensity change before landfall.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Integration of Air Classification and Hydrothermal Carbonization to Enhance Energy Recovery of Corn Stover

Air classification (AC) is a cost-effective technology that separates the energy-dense light ash fraction (LAF) from the inorganic-rich high ash fraction (HAF) of corn stover. HAF could be upgraded into energy-dense solid fuel by hydrothermal carbonization (HTC). However, HTC is a high-temperature, high-pressure process, which requires additional energy to operate. In this study, three different scenarios (i.e., AC only, HTC only, and integrated AC–HTC) were investigated for the energy recovery of corn stover. AC was performed on corn stover at an 8 Hz fan speed, which yielded 84.4 wt. % LAF, 12.8 wt. % HAF, and 2.8 wt. % below screen particles. About 27 wt. % ash was reduced from LAF by the AC process. Furthermore, HTC was performed on raw corn stover and the HAF of corn stover at 200, 230, and 260 °C for 30 min. To evaluate energy recovery, solid products were characterized in terms of mass yield, ash yield, ultimate analysis, proximate analyses, and higher heating value (HHV). The results showed that the energy density was increased with the increase in HTC temperature, meanwhile the mass yield and ash yield were decreased with the increase in HTC temperature. Proximate analysis showed that fixed carbon increased 18 wt. % for original char and 27 wt. % for HAF char at 260 °C, compared to their respective feedstocks. Finally, the hydrochar resulting from HAF was mixed with LAF and pelletized at 180 bar and 90 °C to densify the energy content. An energy balance of the integrated AC–HTC process was performed, and the results shows that integrated AC with HTC performed at 230 °C resulted in an additional 800 MJ/ton of energy recovery compared to the AC-only scenario.

09 BIOMASS FUELS↗
DOE OSTI2021

CyBERT: Cybersecurity Claim Classification by Fine-Tuning the BERT Language Model

We introduce CyBERT, a cybersecurity feature claims classifier based on bidirectional encoder representations from transformers and a key component in our semi-automated cybersecurity vetting for industrial control systems (ICS). To train CyBERT, we created a corpus of labeled sequences from ICS device documentation collected across a wide range of vendors and devices. This corpus provides the foundation for fine-tuning BERT’s language model, including a prediction-guided relabeling process. We propose an approach to obtain optimal hyperparameters, including the learning rate, the number of dense layers, and their configuration, to increase the accuracy of our classifier. Fine-tuning all hyperparameters of the resulting model led to an increase in classification accuracy from 76% obtained with BertForSequenceClassification’s original architecture to 94.4% obtained with CyBERT. Furthermore, we evaluated CyBERT for the impact of randomness in the initialization, training, and data-sampling phases. CyBERT demonstrated a standard deviation of ±0.6% during validation across 100 random seed values. Finally, we also compared the performance of CyBERT to other well-established language models including GPT2, ULMFiT, and ELMo, as well as neural network models such as CNN, LSTM, and BiLSTM. The results showed that CyBERT outperforms these models on the validation accuracy and the F1 score, validating CyBERT’s robustness and accuracy as a cybersecurity feature claims classifier.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Two-Stage Wildlife Event Classification for Edge Deployment

Camera-based wildlife monitoring is often overwhelmed by non-target triggers and slowed by manual review or cloud-dependent inference, which can prevent timely intervention for high stakes human–wildlife conflicts. Our key contribution is a deployable, fully offline edge vision sensor that achieves near-real-time, highly accurate wildlife event classification by combining detector-based empty-image suppression with a lightweight classifier trained with a staged transfer-learning curriculum. Specifically, Stage 1 uses a pretrained You Only Look Once (YOLO)-family detector for permissive animal localization and empty-trigger suppression, and Stage 2 uses a lightweight EfficientNet-based binary classifier to confirm puma on detector crops and gate downstream actions. Our design is robust to low-quality nighttime monochrome imagery (motion blur, low contrast, illumination artifacts, and partial-body captures) and operates using commercially available components in connectivity-limited settings. In field deployments running since May 2025, end-to-end latency from camera trigger to action command is approximately 4 s. Ablation studies using a dataset of labeled wildlife images (pumas, not pumas) show that the two-stage approach substantially reduces false alarms in identifying pumas relative to a full-image classifier while maintaining high recall. On the held-out test set (N = 1434 events), the proposed two-stage cascade achieves precision 0.983, recall 0.975, F1 0.979, accuracy 0.986, and balanced accuracy 0.983, with only 8 false positives and 12 false negatives. The system can be easily adapted for other species, as demonstrated by rapid retraining of the second stage to classify ringtails. Downstream responses (e.g., notifications and optional audio/light outputs) provide flexible actuation capabilities that can be configured to support intervention.

58 GEOSCIENCES↗
DOE OSTI2016

OPTICAL SPECTROSCOPY OF SDSS J004054.65-0915268: THREE POSSIBLE SCENARIOS FOR THE CLASSIFICATION. A z ∼ 5 BL LACERTAE, A BLUE FSRQ, OR A WEAK EMISSION LINE QUASAR

The haunt of high-redshift BL Lacerate objects is day by day more compelling to firmly understand their intrinsic nature and evolution. SDSS J004054.65-0915268 is, at the moment, one of the most distant BL Lac candidates, at z ∼ 5. We present a new optical-near-IR spectrum obtained with ALFOSC-NOT with a new, custom designed dispersive grating aimed to detect broad emission lines that could disprove this classification. In the obtained spectra, we do not detect any emission features and we provide an upper limit to the luminosity of the C iv broad emission line. Therefore, the nature of the object is then discussed, building the overall spectral energy distribution (SED) and fitting it with three different models. Our fits, based on SED modeling with different possible scenarios, cannot rule out the possibility that this source is indeed a BL Lac object, though the absence of optical variability and the lack of strong radio flux seem to suggest that the observed optical emission originates from a thermalized accretion disk.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2016

SN REFSDAL: CLASSIFICATION AS A LUMINOUS AND BLUE SN 1987A-LIKE TYPE II SUPERNOVA

We have acquired Hubble Space Telescope (HST) and Very Large Telescope near-infrared spectra and images of supernova (SN) Refsdal after its discovery as an Einstein cross in fall 2014. The HST light curve of SN Refsdal has a shape consistent with the distinctive, slowly rising light curves of SN 1987A-like SNe, and we find strong evidence for a broad H α P-Cygni profile and Na I D absorption in the HST grism spectrum at the redshift ( z = 1.49) of the spiral host galaxy. SNe IIn, largely powered by circumstellar interaction, could provide a good match to the light curve of SN Refsdal, but the spectrum of a SN IIn would not show broad and strong H α and Na I D absorption. From the grism spectrum, we measure an H α expansion velocity consistent with those of SN 1987A-like SNe at a similar phase. The luminosity, evolution, and Gaussian profile of the H α emission of the WFC3 and X-shooter spectra, separated by ∼2.5 months in the rest frame, provide additional evidence that supports the SN 1987A-like classification. In comparison with other examples of SN 1987A-like SNe, photometry of SN Refsdal favors bluer B - V and V - R colors and one of the largest luminosities for the assumed range of potential magnifications. The evolution of the light curve at late times will provide additional evidence about the potential existence of any substantial circumstellar material. Using MOSFIRE and X-shooter spectra, we estimate a subsolar host-galaxy metallicity (8.3 ± 0.1 dex and <8.4 dex, respectively) near the explosion site.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2016

The K2 Ecliptic Plane Input Catalog (EPIC) and stellar classifications of 138,600 targets in campaigns 1–8

The K2 Mission uses the Kepler spacecraft to obtain high-precision photometry over ≈80 day campaigns in the ecliptic plane. The Ecliptic Plane Input Catalog (EPIC) provides coordinates, photometry, and kinematics based on a federation of all-sky catalogs to support target selection and target management for the K2 mission. We describe the construction of the EPIC, as well as modifications and shortcomings of the catalog. Kepler magnitudes (Kp) are shown to be accurate to ≈0.1 mag for the Kepler field, and the EPIC is typically complete to Kp ≈ 17 (Kp ≈ 19 for campaigns covered by Sloan Digital Sky Survey). We furthermore classify 138,600 targets in Campaigns 1–8 (≈88% of the full target sample) using colors, proper motions, spectroscopy, parallaxes, and galactic population synthesis models, with typical uncertainties for G-type stars of ≈3% in T{sub eff}, ≈0.3 dex in log g, ≈40% in radius, ≈10% in mass, and ≈40% in distance. Our results show that stars targeted by K2 are dominated by K–M dwarfs (≈41% of all selected targets), F–G dwarfs (≈36%), and K giants (≈21%), consistent with key K2 science programs to search for transiting exoplanets and galactic archeology studies using oscillating red giants. However, we find significant variation of the fraction of cool dwarfs with galactic latitude, indicating a target selection bias due to interstellar reddening and increased contamination by giant stars near the galactic plane. We discuss possible systematic errors in the derived stellar properties, and differences with published classifications for K2 exoplanet host stars. The EPIC is hosted at the Mikulski Archive for Space Telescopes (MAST): http://archive.stsci.edu/k2/epic/search.php.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

An Optical Overview of Blazars with LAMOST. II. Gamma-Ray Blazar Candidates and Updated Classifications

Blazars represent the dominant class of associated γ-ray sources detected by the Fermi Large Area Telescope (LAT). However, in the more recent release of the Fourth Fermi-LAT Point Source Catalog (4FGL), ∼25% of the sources associated with lower-energy counterparts show a multifrequency behavior similar to that of blazars, but lacks an optical spectroscopic confirmation of their nature and are therefore classified as Blazar Candidates of Uncertain Type (BCUs). A particularly challenging task in blazar studies is to classify these BCUs and, when possible to estimate their redshifts, in particular for BL Lac objects, characterized by almost featureless optical spectra with only weak emission lines. Continuing our study of blazars with Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) optical spectral data, we perform an extensive search for optical spectra available in the LAMOST Data Release 6 archive. Our aim is confirming the blazar nature of BCUs and to test if new data can allow us to get a redshift estimate for BL Lac objects that lack measurement, as well as to search for and discover changing-look blazars. We selected sources out of the 4FGL catalog, the list of targets from our follow-up spectroscopic campaign of unidentified and/or unassociated γ-ray sources, and the Roma-BZCAT multifrequency catalog of blazars, finding a total of 42 sources with available LAMOST DR6 spectra. We confirmed the blazar-like nature of four blazar candidates . For the remaining 37 sources we confirm their previous classification.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2017

Evaluating Optical Classification for Fermi Blazar Candidates with a Statistical Method Using Broadband Spectral Indices

We aim to test whether a blazar candidate of uncertain type (BCU) in the third Fermi catalog of active galactic nuclei (3LAC) can be potentially classified as a BL Lac object or a flat-spectrum radio quasar (FSRQ) by performing a statistical analysis of its broadband spectral properties. We find that 34% of the radio-selected BCUs (583 BCUs) are BL Lac-like and 20% are FSRQ-like, at a 90% level of confidence. Similarly, 77.3% of the X-ray-selected BCUs (176 BCUs) are evaluated as BL Lac-like and 6.8% may be FSRQ-like sources. And 88.7% of the BL Lac-like BCUs that have synchrotron peak frequencies available are high synchrotron peaked BL Lacs in the X-ray-selected BCUs. The percentages are accordingly 62% and 7.3% in the sample of 124 optically selected BCUs. The high ratio of the number of BL Lac-like sources to the number of FSRQ-like BCUs in the X-ray-selected and optically selected BCU samples is due to selection effects. Examining the consistency between our evaluation and spectroscopic identification case by case with a sample of 78 radio-selected BCUs, it is found that the statistical analysis and its resulting classifications agree with the results of the optical follow-up spectroscopic observations. Our observation campaign for high-|ρ{sub s}| BCUs selected with our method, i.e., |ρ{sub s}|>0.8, is ongoing.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2019

Unblinded Data for PLAsTiCC Classification Challenge

For classification challenge (https://plasticc.org), the unblinded data files are included here. See PDF note above for more information. The original challenge (Sep 28, 2018 - Dec 17, 2018) was hosted at https://www.kaggle.com/c/PLAsTiCC-2018 .

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Data supporting manuscript from L. Sheneman, G. Stephanopoulos, A.E. Vasdekis titled "Deep learning classification of lipid droplets in quantitative phase images" as currently under review at PLOS ONE. This includes: 1) raw and binary labeled Quantitative Phase Images (QPI) of Y. lipolytica cells used in the analyses described within the manuscript. 2) various derived data including classifier scores, etc.

Data supporting manuscript from L. Sheneman, G. Stephanopoulos, A.E. Vasdekis titled "Deep learning classification of lipid droplets in quantitative phase images" as currently under review at PLOS ONE. This includes: 1) raw and binary labeled Quantitative Phase Images (QPI) of Y. lipolytica cells used in the analyses described within the manuscript. 2) various derived data including classifier scores, etc.

ANN↗
DOE OSTI2010

Error Detection and Error Classification: Failure Awareness in Data Transfer Scheduling

Data transfer in distributed environment is prone to frequent failures resulting from back-end system level problems, like connectivity failure which is technically untraceable by users. Error messages are not logged efficiently, and sometimes are not relevant/useful from users point-of-view. Our study explores the possibility of an efficient error detection and reporting system for such environments. Prior knowledge about the environment and awareness of the actual reason behind a failure would enable higher level planners to make better and accurate decisions. It is necessary to have well defined error detection and error reporting methods to increase the usability and serviceability of existing data transfer protocols and data management systems. We investigate the applicability of early error detection and error classification techniques and propose an error reporting framework and a failure-aware data transfer life cycle to improve arrangement of data transfer operations and to enhance decision making of data transfer schedulers.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2015

Impact of Passive Safety on FHR Instrumentation Systems Design and Classification

Fluoride salt-cooled high-temperature reactors (FHRs) will rely more extensively on passive safety than earlier reactor classes. 10CFR50 Appendix A, General Design Criteria for Nuclear Power Plants, establishes minimum design requirements to provide reasonable assurance of adequate safety. 10CFR50.69, Risk-Informed Categorization and Treatment of Structures, Systems and Components for Nuclear Power Reactors, provides guidance on how the safety significance of systems, structures, and components (SSCs) should be reflected in their regulatory treatment. The Nuclear Energy Institute (NEI) has provided 10 CFR 50.69 SSC Categorization Guideline (NEI-00-04) that factors in probabilistic risk assessment (PRA) model insights, as well as deterministic insights, through an integrated decision-making panel. Employing the PRA to inform deterministic requirements enables an appropriately balanced, technically sound categorization to be established. No FHR currently has an adequate PRA or set of design basis accidents to enable establishing the safety classification of its SSCs. While all SSCs used to comply with the general design criteria (GDCs) will be safety related, the intent is to limit the instrumentation risk significance through effective design and reliance on inherent passive safety characteristics. For example, FHRs have no safety-significant temperature threshold phenomena, thus enabling the primary and reserve reactivity control systems required by GDC 26 to be passively, thermally triggered at temperatures well below those for which core or primary coolant boundary damage would occur. Moreover, the passive thermal triggering of the primary and reserve shutdown systems may relegate the control rod drive motors to the control system, substantially decreasing the amount of safety-significant wiring needed. Similarly, FHR decay heat removal systems are intended to be running continuously to minimize the amount of safety-significant instrumentation needed to initiate operation of systems and components important to safety as required in GDC 20. This paper provides an overview of the design process employed to develop a pre-conceptual FHR instrumentation architecture intended to lower plant capital and operational costs by minimizing reliance on expensive, safety related, safety-significant instrumentation through the use of inherent passive features of FHRs.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2017

Quantification of functional impact classification on the current U.S. nuclear fleet

Currently, nuclear power generates only a small fraction of the world's electricity, though it does have the potential to meet the needs of the world as long as its production can compete with the alternative energy base load producers. In U.S the commercial nuclear fleet comprises of Light Water Reactors (LWRs), consisting of Pressurized Water Reactors (PWRs) and Boiling Water Reactors (BWRs). Many plants experience multiple events each year, which lead to reactor trips or scrams. This subsequently leads to plants being off the grid, which is not an economically favorable situation in the current energy markets. This study helps identify functional impact (FI) event categories for both PWR's and BWR's. Understanding that not all events and their causes can be anticipated, current industry best practices implement lessons learned to provide challenges and issues identification. In turn these are studied to achieve improved operational practices, and design modifications. This leads to overall enhanced resiliency. FI groups are the categories of initiating events (IE). IEs are unplanned events that occur while a nuclear plant is in critical operation, and requires that the plant be shut down to achieve a stable state. When a risk cannot be eliminated, IEs are identified and studied to reduce the risk and contribute to the identification of mitigations, safer designs and operational practices. Breaking IE events into groups' results in the classification of FI groups. A FI group is a risk-significant event category that could impact the ability of a NPP to remove decay heat from the reactor. Inability to remove decay heat from the reactor has the potential to lead to core damage. For an event to be considered a FI, it must be associated with a manual or automatic reactor trip, irrespective of the order of events. In addition, the plant must be at or above the point of adding heat and the event must happen before or shortly after a reactor trip. One or more FIs may be identified in any individual reactor trip sequence. The relationship between FI sequence, and/or frequency per a plant event is outside the scope of this paper. For example, the loss of offsite power (LOOP) can occur from eternal events (i.e. weather, human, wildlife etc.). This results in an event that requires a transfer of power to an emergency source, such as a generator, which produces a reactor trip and then a closure of main steam isolation valves (MSIVS). The FI applicable to this reactor trip sequence would be LOOP and closure of at least one MSIV in each main steam line. The results presented herein are focused on the analysis and evaluation of FI event categories. Only FI categories that can be found in IE studies, Initialing Events Spreadsheet and are recorded in a historical review of Licensee Event Reports (LERs) from 1988 through 2013. The results of this analysis can support plant decisions, design, procedures, and identify weakness.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Robustness of Deep Learning Classification to Adversarial Input on GPUs: Asynchronous Parallel Accumulation Is a Source of Vulnerability

The ability of machine learning (ML) classification models to resist small, targeted input perturbations—known as adversarial attacks—is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets used in robustness assessments. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.

Shanmugavelu, Sanjif [Maxeler Technologies, a Groq↗