Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Labeling”

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 523 records · Page 29
All sourcesNASA NTRSDOE OSTI
NASA NTRS1992

Development and use of domain-specific antibodies in a characterization of the large subunits of soybean photosystem 1

The molecular architecture of the soybean photosystem 1 reaction center complex was examined using a combination of surface labeling and immunological methodology on isolated thylakoid membranes. Synthetic peptides (12 to 14 amino acids in length) were prepared which correspond to the N-terminal regions of the 83 and 82.4 kDa subunits of photosystem 1 (the PsaA and PsaB proteins, respectively). Similarly, a synthetic peptide was prepared corresponding to the C-terminal region of the PsaB subunit. These peptides were conjugated to a carrier protein, and were used for the production of polyclonal antibodies in rabbits. The resulting sera could distinguish between the PsaA and PsaB photosystem 1 subunits by Western blot analysis, and could identify appropriate size classes of cyanogen bromide cleavage fragments as predicted from the primary sequences of these two subunits. When soybean thylakoid membranes were surface-labeled with N-hydroxysuccinimidobiotin, several subunits of the complete photosystem 1 lipid/protein complex incorporated label. These included the light harvesting chlorophyll proteins of photosystem 1, and peptides thought to aid in the docking of ferredoxin to the complex during photosynthetic electron transport. However, the PsaA and PsaB subunits showed very little biotinylation. When these subunits were examined for the domains to which biotin did attach, most of the observed label was associated with the N-terminal domain of the PsaA subunit, as identified using a domain-specific polyclonal antisera.

NASA Program NSCORT↗
NASA NTRS1993

Characterization of auxin-binding proteins from zucchini plasma membrane

We have previously identified two auxin-binding polypeptides in plasma membrane (PM) preparations from zucchini (Cucurbita pepo L.) (Hicks et al. 1989, Proc. Natl. Acad. Sci. USA 86, 4948-4952). These polypeptides have molecular weights of 40 kDa and 42 kDa and label specifically with the photoaffinity auxin analog 5-N3-7-3H-IAA (azido-IAA). Azido-IAA permits both the covalent and radioactive tagging of auxin-binding proteins and has allowed us to characterize further the 40-kDa and 42-kDa polypeptides, including the nature of their attachment to the PM, their relationship to each other, and their potential function. The azido-IAA-labeled polypeptides remain in the pelleted membrane fraction following high-salt and detergent washes, which indicates a tight and possibly integral association with the PM. Two-dimensional electrophoresis of partially purified azido-IAA-labeled protein demonstrates that, in addition to the major isoforms of the 40-kDa and 42-kDa polypeptides, which possess isoelectric points (pIs) of 8.2 and 7.2, respectively, several less abundant isoforms that display unique pIs are apparent at both molecular masses. Tryptic and chymotryptic digestion of the auxin-binding proteins indicates that the 40-kDa and 42-kDa polypeptides are closely related or are modifications of the same polypeptide. Phase extraction with the nonionic detergent Triton X-114 results in partitioning of the azido-IAA-labeled polypeptides into the aqueous (hydrophilic) phase. This apparently paradoxical behavior is also exhibited by certain integral membrane proteins that aggregate to form channels. The results of gel filtration indicate that the auxin-binding proteins do indeed aggregate strongly and that the polypeptides associate to form a dimer or multimeric complex in vivo. These characteristics are consistent with the hypothesis that the 40-kDa and 42-kDa polypeptides are subunits of a multimeric integral membrane protein which has an auxin-binding site, and which may possess transporter or channel function.

NASA Program Space Biology↗
NASA NTRS1985

Quantitative determination of microbial activity and community nutritional status in estuarine sediments: evidence for a disturbance artifact

In estuarine sediments with a high degree of vertical heterogeneity in reduced substrate and terminal electron acceptor concentrations, the method of exposure of the microbiota to labeled substrates can introduce a "disturbance artifact" into measures of metabolic activity. The detection of this artifact is based on quantitative measurement of the relative rates of incorporation of [14C]acetate into phospholipid fatty acids (PLFA) and endogenous storage lipid, poly-beta-hydroxyalkanoate (PHA). Previous studies have shown that PLFA synthesis measures cellular growth and that PHA synthesis measures carbon accumulation (unbalanced growth). The "disturbance artifact" of exposure to [14C]acetate was demonstrated by comparing injection of a core with the usual or pore-water replacement or slurry techniques. Only injection of labeled substrate allowed detection of preassay disturbance of the sediment with a garden rake. The raking increased PLFA synthesis with little effect to differences in concentration or distribution of [14C]acetate in the 10-min incubation. Bioturbation induced by sand dollar feeding in estuarine sediment could be detected in an increased PLFA/PHA ratio which was due to decreased PHA synthesis if the addition of labeled substrate was by the injection technique. Addition of labeled precursors to sediment by slurry or pore-water replacement induces greater disturbance artifacts than injection techniques.

Phospholipids/analysis/biosynthesis↗
NASA NTRS1999

Incomplete segregation of endorgan-specific vestibular ganglion cells in mice and rats

The endorgan-specific distribution of vestibular ganglion cells was studied in neonatal and postnatal rats and mice using indocarbocyanine dye (DiI) and dextran amines for retrograde and anterograde labeling. Retrograde DiI tracing from the anterior vertical canal labeled neurons scattered throughout the whole superior vestibular ganglion, with denser labeling at the dorsal and central regions. Horizontal canal neurons were scattered along the dorsoventral axis with more clustering toward the dorsal and ventral poles of this axis. Utricular ganglion cells occupied predominantly the central region of the superior vestibular ganglion. This utricular population overlapped with both the anterior vertical and horizontal canals' ganglion cells. Posterior vertical canal neurons were clustered in the posterior part of the inferior vestibular ganglion. The saccular neurons were distributed in the two parts of the vestibular ganglion, the superior and inferior ganglia. Within the inferior ganglion, the saccular neurons were clustered in the anterior part. In the superior ganglion, the saccular neurons were widely scattered throughout the whole ganglion with more numerous neurons at the posterior half. Small and large neurons were labeled from all endorgans. Examination of the fiber trajectory within the superior division of the vestibular nerve showed no clear lamination of the fibers innervating the different endorgans. These results demonstrate an overlapping pattern between the different populations within the superior ganglion, while in the inferior ganglion, the posterior canal and saccular neurons show tighter clustering but incomplete segregation. This distribution implies that the ganglion cells are assigned for their target during development in a stochastic rather than topographical fashion.

NASA Discipline Developmental Biology↗
NASA NTRS2006

Fluorescent Approaches to High Throughput Crystallography

We have shown that by covalently modifying a subpopulation, less than or equal to 1%, of a macromolecule with a fluorescent probe, the labeled material will add to a growing crystal as a microheterogeneous growth unit. Labeling procedures can be readily incorporated into the final stages of purification, and the presence of the probe at low concentrations does not affect the X-ray data quality or the crystallization behavior. The presence of the trace fluorescent label gives a number of advantages when used with high throughput crystallizations. The covalently attached probe will concentrate in the crystal relative to the solution, and under fluorescent illumination crystals show up as bright objects against a dark background. Non-protein structures, such as salt crystals, will not incorporate the probe and will not show up under fluorescent illumination. Brightly fluorescent crystals are readily found against less bright precipitated phases, which under white light illumination may obscure the crystals. Automated image analysis to find crystals should be greatly facilitated, without having to first define crystallization drop boundaries as the protein or protein structures is all that shows up. Fluorescence intensity is a faster search parameter, whether visually or by automated methods, than looking for crystalline features. We are now testing the use of high fluorescence intensity regions, in the absence of clear crystalline features or "hits", as a means for determining potential lead conditions. A working hypothesis is that kinetics leading to non-structured phases may overwhelm and trap more slowly formed ordered assemblies, which subsequently show up as regions of brighter fluorescence intensity. Preliminary experiments with test proteins have resulted in the extraction of a number of crystallization conditions from screening outcomes based solely on the presence of bright fluorescent regions. Subsequent experiments will test this approach using a wider range of proteins. The trace fluorescently labeled crystals will also emit with sufficient intensity to aid in the automation of crystal alignment using relatively low cost optics, further increasing throughput at synchrotrons.

Pusey, Marc L.↗
NASA NTRS2004

Active learning in the presence of unlabelable examples

We propose a new active learning framework where the expert labeler is allowed to decline to label any example. This may be necessary because the true label is unknown or because the example belongs to a class that is not part of the real training problem. We show that within this framework, popular active learning algorithms (such as Simple) may perform worse than random selection because they make so many queries to the unlabelable class. We present a method by which any active learning algorithm can be modified to avoid unlabelable examples by training a second classifier to distinguish between the labelable and unlabelable classes. We also demonstrate the effectiveness of the method on two benchmark data sets and a real-world problem.

active learning↗
NASA NTRS2006

Enhancement of the MODIS Snow and Ice Product Suite Utilizing Image Segmentation

A problem has been noticed with the current NODIS Snow and Ice Product in that fringes of certain snow fields are labeled as "cloud" whereas close inspection of the data indicates that the correct labeling is a non-cloud category such as snow or land. This occurs because the current MODIS Snow and Ice Product generation algorithm relies solely on the MODIS Cloud Mask Product for the labeling of image pixels as cloud. It is proposed here that information obtained from image segmentation can be used to determine when it is appropriate to override the cloud indication from the cloud mask product. Initial tests show that this approach can significantly reduce the cloud "fringing" in modified snow cover labeling. More comprehensive testing is required to determine whether or not this approach consistently improves the accuracy of the snow and ice product.

Tilton, James C.↗
NASA NTRS2008

User Guide for the Anvil Threat Cooridor Forecast Tool V2.4 for AWIPS

The Anvil Tool GUI allows users to select a Data Type, toggle the map refresh on/off, place labels, and choose the Profiler Type (source of the KSC 50 MHz profiler data), the Date- Time of the data, the Center of Plot, and the Station (location of the RAOB or 50 MHz profiler). If the Data Type is Models, the user selects a Fcst Hour (forecast hour) instead of Station. There are menus for User Profiles, Circle Label Options, and Frame Label Options. Labels can be placed near the center circle of the plot and/or at a specified distance and direction from the center of the circle (Center of Plot). The default selection for the map refresh is "ON". When the user creates a new Anvil Tool map with Refresh Map "ON, the plot is automatically displayed in the AWIPS frame. If another Anvil Tool map is already displayed and the user does not change the existing map number shown at the bottom of the GUI, the new Anvil Tool map will overwrite the old one. If the user turns the Refresh Map "OFF", the new Anvil Tool map is created but not automatically displayed. The user can still display the Anvil Tool map through the Maps dropdown menu* as shown in Figure 4.

Barett, Joe H., III↗
NASA NTRS2015

Johnson Space Center's Free Range Bicycle Program.- Fall 2015 Intern Report

NASA's Johnson Space Center is a big place, encompassing 1,620 acres and more than a hundred buildings. Furthermore, there are reportedly 15 thousand employees, all of which have somewhere to be. To facilitate the movement of all these people JSC has historically relied on human power. Pedaling their way towards deep space, bicycles have been the go to method. Currently there are about 200 Free Range Bicycles at JSC. Free Range Bicycles belong to nobody, except NASA, and are available for anybody to use. They are not to be locked or hidden (although frequently are) and the intention is that there will always be a bike to hop on to get where you're going (although it may not be the bike you rode in on). Although not without its own shortcomings, the Free Range Bicycle Program has continued to provide low cost, simple transportation for NASA's JSC. In addition to the approximately 200 Free Range Bicycles, various larger divisions (like engineering) will often buy a few dozen bikes for their team members to use or individuals will bring their own personal bike to either commute or use on site. When these bicycles fall into disrepair or are abandoned (from retirees etc) they become a problem at JSC. They are an eye sore, create a safety hazard and make it harder to find a working bike in a time of need. The Free Range Program hopes to address this first problem by "tagging out" abandoned or out of service bicycles. A bright orange "DO NOT OPERATE" tag is placed on the bike and given a serial number for tracking purposes. See picture to the right. If the bike has an active owner with intentions to repair the bike the bottom of the tag has instructions for how to claim the abandoned bicycle. After being tagged the owner of the bicycle has 30 days to claim the bicycle and either haul it off site or get it repaired (and labeled) in accordance with Johnson's Bicycle Policy. If the abandoned bicycle is not claimed within 30 days it becomes the property of the Government. The bicycle is then (in short) repaired, labeled, documented and converted to a free range bicycle. Bikes beyond repair are cannibalized of useable parts and then scrapped. That was nearly the first thing I did when arriving at work. After getting settled in and coordinating the purchase of 50 new bicycles (elaboration below); I started combing the Center. Bikes are hidden and tucked away in the oddest of places and it was my priority to root all of them out. I tagged 70 bikes on the center, setting a record. I tagged so many bikes I ran out tags and had to make more. Long ago it was discovered that the same harsh elements that wreak havoc on the bikes, destroy the tags before the 30 day timeframe. The tags are labeled with instructions on how to claim the bike and numbered, then these paper labels are laminated with packing tape. It sounds like a simple process but nothing fits right, everything has to be trimmed, put on straight and is generally a pain in the neck. So I made an assembly process and knocked out a few hundred to save having to do it again for a while. I walked the center and tagged bikes at every building hiding in nearly every nook and cranny. Thus, in conclusion, I've done many things here at JSC on my first term and had a blast doing it. I've learned a lot from my multi-faceted roles and continual challenges. In short I've done my best to get folks at JSC on bikes and keep folks at JSC on bikes and tended a flock of free range bicycles. Everything from turning the rusty bolts and oiling chains to coordinating a $20,000 deal on 50 new free range bicycles. I am proud of the bicycle shop I have built and am infinitely grateful to the people who lead, help, guide and support me. I have fixed a great number of bicycles and cleaned the center of unsightly and unsafe piles of bicycles. I am immensely thankful for this opportunity to learn and serve and appreciate the privilege of coming back for a second term. A second term, in which I will continue to develop this awesome program.

Lee-Stockton, Willem↗
NASA NTRS2017

Ask-the-expert: Active Learning Based Knowledge Discovery Using the Expert

Often the manual review of large data sets, either for purposes of labeling unlabeled instances or for classifying meaningful results from uninteresting (but statistically significant) ones is extremely resource intensive, especially in terms of subject matter expert (SME) time. Use of active learning has been shown to diminish this review time significantly. However, since active learning is an iterative process of learning a classifier based on a small number of SME-provided labels at each iteration, the lack of an enabling tool can hinder the process of adoption of these technologies in real-life, in spite of their labor-saving potential. In this demo we present ASK-the-Expert, an interactive tool that allows SMEs to review instances from a data set and provide labels within a single framework. ASK-the-Expert is powered by an active learning algorithm for training a classifier in the backend. We demonstrate this system in the context of an aviation safety application, but the tool can be adopted to work as a simple review and labeling tool as well, without the use of active learning.

software↗
NASA NTRS2017

Ask-the-Expert: Active Learning Based Knowledge Discovery Using the Expert

Often the manual review of large data sets, either for purposes of labeling unlabeled instances or for classifying meaningful results from uninteresting (but statistically significant) ones is extremely resource intensive, especially in terms of subject matter expert (SME) time. Use of active learning has been shown to diminish this review time significantly. However, since active learning is an iterative process of learning a classifier based on a small number of SME-provided labels at each iteration, the lack of an enabling tool can hinder the process of adoption of these technologies in real-life, in spite of their labor-saving potential. In this demo we present ASK-the-Expert, an interactive tool that allows SMEs to review instances from a data set and provide labels within a single framework. ASK-the-Expert is powered by an active learning algorithm for training a classifier in the back end. We demonstrate this system in the context of an aviation safety application, but the tool can be adopted to work as a simple review and labeling tool as well, without the use of active learning.

GUI↗
NASA NTRS2020

Efficient Active Learning for New Domains

The promise of active learning is to reduce the number of labeled examples required by supervised machine learning algorithms. The largest potential benefits lie in entirely new domains, for which no labeled examples yet exist. Yet to date, most active learning studies are retroactive and demonstrate the benefits that could have been gained if active learning had been used. What are the barriers to true adoption and utilization of active learning? We focus on two: (1) the cold start or class discovery problem, in which active learning methods may struggle to make progress with zero labeled examples, and (2) the cost of having the classifier in the loop to select the next example to be labeled. We assess different active learning approaches in the context of these two barriers and conclude with recommendations for how to employ active learning in new domains. As an example, we report on the use of active learning on a large, novel data set of Mars surface images.

Lu, Steven↗
NASA NTRS2021

AI4MARS: A Dataset for Terrain-Aware Autonomy on Mars

Deep learning has quickly become a necessity for selfdriving vehicles on Earth. In contrast, the self-driving vehicles on Mars, including NASA’s latest rover, Perseverance, which is planned to land on Mars in February 2021, are still driven by classical machine vision systems. Deep learning capabilities, such as semantic segmentation and object recognition, would substantially benefit the safety and productivity of ongoing and future missions to the red planet. To this end, we created the first large-scale dataset, AI4Mars, for training and validating terrain classification models for Mars, consisting of ~326K semantic segmentation full image labels on 35K images from Curiosity, Opportunity, and Spirit rovers, collected through crowdsourcing. Each image was labeled by ~10 people to ensure greater quality and agreement of the crowdsourced labels. It also includes ~1.5K validation labels annotated by the rover planners and scientists from NASA’s MSL (Mars Science Laboratory) mission, which operates the Curiosity rover, and MER (Mars Exploration Rovers) mission, which operated the Spirit and Opportunity rovers. We trained a DeepLabv3 model on the AI4Mars training dataset and achieved over 96% overall classification accuracy on the test set. The dataset is made publicly available.1

Ono, Hiro↗
NASA NTRSDate not supplied

Anomaly Detection in Flight Operational Data Using Deep Learning

In this session, we demonstrate two recently developed deep learning models for anomaly detection in flight operational data by the Data Sciences Group at NASA Ames Research Center. The first model is Convolutional Variational Auto-Encoder (CVAE) [1], which is an unsupervised deep encoder-decoder model, designed specifically for finding anomalies in heterogeneous multivariate time series data. We will demonstrate its application to finding anomalies in streaming data from NASA’s Digital Information Platform’s Fuser source. CVAE identifies data instances that are not representative of expected nominal behavior as anomalous. Since it is an unsupervised approach, the flagged anomalies will need to be reviewed by the subject matter experts (SMEs) for validation and labeling and is designed to assist with vulnerability discovery within Safety Monitoring System programs. The second model is Robust and Explainable Semi-supervised Anomaly Detection (RESAD) model [2], which builds on CVAE to allow learning from both minimally labeled data (previously reviewed by the SMEs) as well as majority unlabeled data. RESAD takes advantage of graph theoretic techniques to propagate the labels from the labeled data to the unlabeled data based on a pre-defined similarity metric and structures the learned feature space from flight time-series so that data of the same class would cluster tightly together. This model characteristic is enabled by training with an augmented loss function and allows learning of a more informative feature space for down-stream tasks such as search and active learning. We demonstrate RESAD using data from the NASA DASHlink project [3].

anomaly detection↗
NASA NTRSDate not supplied

Towards an Aviation Large Language Model by Fine-tuning and Evaluating Transformers

In the aviation domain, there are many applications for machine learning and artificial intelligence tools that utilize natural language. For example, there is a desire to know the commonalities in written safety reports such as voluntary post incidents reports or aerial wildfire operations reports to better understand the risks present. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement. These applications can benefit from the use of state-of-the-art natural language processing techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of adaptation of NLP tools to the aviation domain by fine-tuning transformer based models using aviation data sets. In 2018, a novel language model based on neural units (also called transformers) was created and became known as “Bidirectional Encoder Representations from Transformers” or BERT. This architecture combined with large amounts of English training data and innovative semi-supervised training tasks set the standard for what would later emerge as Large Language Models. The performance of these models was further improved by hyperparameter tuning and refinement of the semi-supervised training task and resulted in “Robustly Optimized BERT Pre-training Approach through hyperparameter tuning” or RoBERTa models. These pre-trained Large Language Models proved to be useful for a wide variety of natural language processing tasks such as text classification and question answering through a process called fine-tuning. The transformer architecture with pre-trained weights served as the basis with the last few layers replaced with layers fine-tuned to perform a new task e.g., a layer that provides a label for the entire input text. This process of fine-tuning can also be used to adapt the models to new domains; e.g., BioBERT started with the pre-trained BERT model and was completed by additional fine-tuning and training on biomedical documents. Transformer-based architectures can also be used to create rich representations of text called embeddings which can serve as the input to other machine learning models. This allows simpler algorithms such as logistic regression to use context-rich representations of the text while still remaining quick to train and evaluate. In the world of aviation, there is a growing demand for natural language processing and understanding but the domain presents unique challenges. Due to the technical content (and specialized language) of most aviation documents, fine-tuning pre-trained Large Language Models to specific tasks has not met the benchmark on natural language processing tasks set by simpler models trained from scratch on the data. To address this deficiency, this paper evaluates the improvements from fine-tuning a Large Language Model on a large set of aviation documents using the original semi-supervised training tasks before performing specific natural language tasks. In fine-tuning, a domain-specific dataset is used on the original training task but with the pre-trained Large Language Model instead of starting from a random initialization. This approach allows the model to be adapted to the specific domain language without discarding the information gained from training on general English data. This paper utilized two major dataset types to train and assess the RoBERTa fine-tuning performance. The first are 7,057 Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the semi-supervised fine-tuning. The second type is the 494 document classification labels to be used for evaluation. This down-stream evaluation aims to show the performance of the fine-tuned model, better understand how much data is needed for an effective fine-tuning, and how fine-tuning can be adapted for different applications in-the domain. After semi-supervised training, evaluation begins by encoding the documents for classification using the fine-tuned RoBERTa model. Then a logistic regression classifier is trained to label the document type and compared against our ground truth labels. This currently leads to a 82.8% accuracy on 10-fold cross validation showing improvement over baseline RoBERTa which achieved 81.0%. We plan to measure the improvements on additional tasks and it is expected that these improvements will lead to more robust models that can tackle the natural language processing challenges present in aviation datasets.

ATM↗
DOE OSTI2021

Incorporating historical data and past analyses for improved tensile property prediction of 9% Cr steel

Data-driven analytical clustering and visualization techniques were applied to the dataset of 9% Cr experimental alloy data generated through the eXtremeMAT project. Techniques and results were compared with the resulting clusters obtained through similar analytical techniques on previous and reduced versions of the dataset. The principal components were generated in order to reduce the dimensionality of the complex dataset and to visualize the underlying trends in the data. Partitioning around medoids was performed on the resulting principal components to determine relevant clusters. Domain knowledge labels were further applied to the principal components to compare the labels with the trends identified through the clustering methods. The resulting clusters show the addition of the new data into the dataset, and show where alignment between the domain knowledge labels and partitioning around medoids (PAM) clusters show where the labels are appropriate. The clusters can be used to compare the tensile properties of the alloys, and to reduce the variation in the dataset.

Alloy design, Mechanical Properties, data analysis↗
DOE OSTI2022

Uncertainty-refined image segmentation under domain shift

A method for digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. The label of any image element with an uncertainty value above a predefined threshold is replaced with a new label corresponding to the class with the next highest value.

Martinez, Carianne↗
DOE OSTI2025

AI-Enabled Operations at Fermi Complex: Multivariate Time Series Prediction for Outage Prediction and Diagnosis

The Main Control Room of the Fermilab accelerator complex continuously gathers extensive time-series data from thousands of sensors monitoring the beam. However, unplanned events such as trips or voltage fluctuations often result in beam outages, causing operational downtime. This downtime not only consumes operator effort in diagnosing and addressing the issue but also leads to unnecessary energy consumption by idle machines awaiting beam restoration. The current threshold-based alarm system is reactive and faces challenges including frequent false alarms and inconsistent outage-cause labeling. To address these limitations, we propose an AI-enabled framework that leverages predictive analytics and automated labeling. Using data from $2,703$ Linac devices and $80$ operator-labeled outages, we evaluate state-of-the-art deep learning architectures, including recurrent, attention-based, and linear models, for beam outage prediction. Additionally, we assess a Random Forest-based labeling system for providing consistent, confidence-scored outage annotations. Our findings highlight the strengths and weaknesses of these architectures for beam outage prediction and identify critical gaps that must be addressed to fully harness AI for transitioning downtime handling from reactive to predictive, ultimately reducing downtime and improving decision-making in accelerator management.

Jain, Milan [PNL, Richland] (ORCID:000000021676111↗