Search NASA⌕ Search

SEARCH · Search NASA

Results for “unsupervised learning”

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.

251 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Forming a database to study reversed magnetic shear from the National Spherical Torus eXperiment using machine learning

Achieving a long-lived reversed magnetic shear (RMS) target plasma in the National Spherical Torus eXperiment Upgrade will require developing various sustainment scenarios. To help with the ongoing plasma control efforts, the development of a new analysis for the motional Stark effect (MSE) diagnostic using a machine learning algorithm, namely, MSE-ML, is described. MSE-ML will be used to identify patterns during RMS discharges, some of which suffer magnetohydrodynamic (MHD) events resulting in current redistribution and monotonic q-profiles. A database consisting of q and magnetic shear profiles is being constructed primarily based on the existing National Spherical Torus eXperiment data with equilibrium reconstructions constrained by the magnetic field pitch angle profile measured using the multi-channel MSE diagnostic. An unsupervised k-means clustering of the data is developed to study the RMS formation as a function of time. The initial clustering from the q-profiles shows significant differences in both amplitude and the duration of the RMS period. As a goal, the clustering results that detect and distinguish shots with substantial and sustained RMS are to be used as a preprocessing step in a supervised algorithm to identify the underlying conditions that lead to long-lasting improved confinement with RMS. Another aim of the MSE-ML study is to identify precursors of RMS-destroying MHD events in either derived data such as the q-profile or directly measured data such as the magnetic field pitch angle profile.

Uzun-Kaymak, I. U. (ORCID:0000000276251493)↗
DOE OSTI2024

Anomaly detection with flow-based fast calorimeter simulators

Recently, several normalizing flow-based deep generative models have been proposed to accelerate the simulation of calorimeter showers. Using caloflow as an example, we show that these models can simultaneously perform unsupervised anomaly detection with no additional training cost. As a demonstration, we consider electromagnetic showers initiated by one (background) or multiple (signal) photons. The caloflow model is designed to generate single-photon showers, but it also provides access to the shower likelihood. We use this likelihood as an anomaly score and study the showers tagged as being unlikely. As expected, the tagger struggles when the signal photons are nearly collinear but is otherwise effective. This approach is complementary to a supervised classifier trained on only specific signal models using the same low-level calorimeter inputs. While the supervised classifier is also highly effective at unseen signal models, the unsupervised method is more sensitive in certain regions, and, thus, we expect that the ultimate performance will require a combination of these approaches.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Neural Network Prediction of Strong Lensing Systems with Domain Adaptation and Uncertainty Quantification

Modeling strong gravitational lenses is computationally expensive for the complex data from modern and next-generation cosmic surveys. Deep learning has emerged as a promising approach for finding lenses and predicting lensing parameters, such as the Einstein radius. Mean-variance Estimators (MVEs) are a common approach for obtaining aleatoric (data) uncertainties from a neural network prediction. However, neural networks have not been demonstrated to perform well on out-of-domain target data successfully - e.g., when trained on simulated data and applied to real, observational data. In this work, we perform the first study of the efficacy of MVEs in combination with unsupervised domain adaptation (UDA) on strong lensing data. The source domain data is noiseless, and the target domain data has noise mimicking modern cosmology surveys. We find that adding UDA to MVE increases the accuracy on the target data by a factor of about two over an MVE model without UDA. Including UDA also permits much more well-calibrated aleatoric uncertainty predictions. Advancements in this approach may enable future applications of MVE models to real observational data.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Finding Real Uncertainties From Physical Simulations

Modeling strong gravitational lenses is computationally expensive for the complex data from modern and next-generation cosmic surveys. Deep learning has emerged as a promising approach for finding lenses and predicting lensing parameters, such as the Einstein radius. Mean-variance Estimators (MVEs) are a common approach for obtaining aleatoric (data) uncertainties from a neural network prediction. However, neural networks have not been demonstrated to perform well on out-of-domain target data successfully - e.g., when trained on simulated data and applied to real, observational data. In this work, we perform the first study of the efficacy of MVEs in combination with unsupervised domain adaptation (UDA) on strong lensing data. The source domain data is noiseless, and the target domain data has noise mimicking modern cosmology surveys. We find that adding UDA to MVE increases the accuracy on the target data by a factor of about two over an MVE model without UDA. Including UDA also permits much more well-calibrated aleatoric uncertainty predictions. Advancements in this approach may enable future applications of MVE models to real observational data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2026

Rolling Root Mean Square Based Multimodal Anomaly Detection for Real Time Monitoring of Smart Grid

Reliable real-time monitoring is valuable for maintaining the operational integrity of modern electrical smart grids. Deployment of heterogeneous sensing technologies in substations has enabled high-resolution, multichannel waveform monitoring, but also introduces challenges for anomaly detection due to noise, baseline drift, and modality-dependent signal characteristics. In this work, we present a computationally efficient unsupervised method for multimodal event detection based on Rolling Root Mean Square based Event Detection (RRMSED). The method is developed using in-house, field deployed sensors collecting data at a utility substation. The sensing system comprises voltage and current sensors, triaxial accelerometers, and magnetometers, collectively capturing electrical, vibrational, and magnetic waveform measurements at high temporal resolution. RRMSED operates by extracting rolling RMS energy features and their first-order temporal differences from consecutive waveform segments for each channel and then applying channel-specific statistical thresholds learned from historical data. A persistence-based exceedance logic is employed to robustly identify transient events while suppressing impulsive noise, and to provide precise temporal localization with high resolution. The framework is designed for continuous server-side operation and can be deployed in real time without requiring complex models. Experiments on simulated waveform data with known ground truth demonstrate low false positive (FP) and false negative (FN) rates. Application to real substation data shows RRMSED to identify events that are not captured by conventional monitoring indicators including fast transient detection algorithm currently deployed in the system. These results indicate that rolling RMS based features provide an effective and practical basis for real-time multimodal event detection in smart-grid substations.

Mukherjee, Subrata [ORNL] (ORCID:0000000309930338)↗
DOE OSTI2024

SSTDR and FDR Detection of Un-Energized and Energized Cable Anomalies Including Thermal Degradation Using Machine Learning

Historically, cables are initially qualified for nuclear power plant use for 40 years. As plants extend their operating license to 60 and 80 years, continued use of these cables must shift to a performance-based approach since it is cost prohibitive to completely replace cables that are likely still capable of performing their design function. A variety of cable tests are available and are commonly applied during outages when the cables can be taken out of service. Frequency domain reflectometry (FDR) is one of these test methods that is being more broadly accepted and used because it not only detects anomalies along the cable with a low-voltage signal that does not stress the cable insulation, but the technique also locates the anomalies. This supports follow-up local inspection and local repair or partial replacement of a damaged cable segment. Currently, FDR testing is only applied to cables that are taken out of service since the test instrument would be damaged by operational voltages. A related technology that has found some acceptance in the aircraft and rail industry is spread spectrum time domain reflectometry (SSTDR). This technology has been implemented with a custom commercial instrument by LiveWire Innovation that is designed to operate on live cables up to 1000 volts and with a bandwidth of 48 MHz. Initial evaluation by the Pacific Northwest National Laboratory (PNNL) of the Live Wire system indicated that a broader bandwidth (BW) SSTDR may be better for many kinds of flaws. This led PNNL to develop an SSTDR laboratory instrument suitable for tests up to 500 MHz bandwidth. Testing on energized cables is also desirable for online monitoring systems so an inductive clamshell coupler was developed that allows energized cables to be tested up to at least 5 kV and likely higher voltage levels. Dielectric spectroscopy and tan delta testing plus various laboratory destructive tests were included in this data acquisition campaign directed to feed a machine learning (ML) study. With these kinds of developments, online energized cable tests may be possible with industrial adoption of such hardware advances but it will be completely impractical to have highly skilled data analysts continually examine these complex signals for indications of damage or compromised conditions. If online testing is to be implemented in new test hardware, it must be accompanied by software that can interpret the signals and alert plant operators of changing or degraded conditions. The thermally aged, shielded cable investigated here was separately treated for ML analysis. Visual analysis of electrical data showed generally increasing peaks where the cable entered and exited the oven. These peaks were not exactly aligned with expected locations, but these differences were attributed to velocity of propagation calibration errors. Only supervised ML was applied to the thermally aged data as this data was only available shortly before the committed publication date of this report. The supervised ML was structured to divide the 0 to 70-day responses as ‘normal’ from 0 to 35 days or ‘anomalous’ from 36 to 70 days, based on cable tensile elongation at break (EAB) insulation characterization. Using 80% of the data for training and 20% for testing, the supervised ML predicted normal versus anomalous was 70% accurate. Important conclusions include: • Accuracy to predict the presence of cable damage is improved from the 2023 effort by more training data. Weighted accuracies for comparisons among the instruments ranged from 67 to 89 % for unsupervised ML and 71 to 99% for supervised ML. • Based on the synthetic data tests, the unsupervised models are more generalizable to unseen anomalies. The Multi-Layer Perceptron classifier (MLP) model reported as high as 99.7% accuracy on the test data, but this dropped to 58.3% when tested on the synthetic data. In contrast, the unsupervised Pointwise model only achieved 89.7% accuracy on the experimental data but reported 78.3% accuracy on the synthetic data. • The best anomaly indicators are higher frequency (400 MHz BW) FDR data. Other tests may be interesting but for this study, this was the best predicter.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

LeaPP: Learning Pathways to Polymorphs through Machine Learning Analysis of Atomic Trajectories

Understanding the mechanisms underlying crystal nucleation and growth is crucial for many technological applications. Due to the short length and time scales involved, crystal nucleation is often studied using molecular simulations. Most existing approaches to extract the nucleation mechanism from simulations focus on the analysis of static snapshots of the configurations, potentially overlooking subtle local fluctuations and the history of the particles involved in the formation of solid nuclei. Here, in this work, we propose a novel methodology called LeaPP that categorizes nucleation trajectories based on the temporal information of their constituent particles. We leverage the time evolution of the local environment of the crystallizing particles to encapsulate the relationship between the structure and dynamics and distinguish between different evolving particle paths. Identification of the distinct particle paths further enables characterizing the nucleation trajectories into different pathways. Collectively, LeaPP provides a more nuanced understanding of nucleation through an unsupervised approach with lesser dependence on traditional order parameters. Furthermore, the pathways identified by LeaPP are predictive of the resulting polymorph. We demonstrate LeaPP on three different systems─Lennard-Jones-like particles, Ni 3 Al, and water on surfaces. The general methodology underlying LeaPP─considering the time evolution of the building blocks─applies to a wide range of self-assembly problems.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Robust Spectral Anomaly Detection in EELS Spectral Images via 3D Convolutional Variational Autoencoders

Abstract A 3D Convolutional Variational Autoencoder (3D‐CVAE) is introduced for automated anomaly detection in electron energy‐loss spectroscopy spectrum imaging (EELS‐SI) data. This approach leverages the full 3D structure of EELS‐SI data to detect subtle spectral anomalies while preserving both spatial and spectral correlations across the datacube. By employing cross‐entropy loss and training on bulk spectra, the model learns to reconstruct bulk features characteristic of the defect‐free material. In exploring methods for anomaly detection, both the 3D‐CVAE approach and principal component analysis (PCA) are evaluated, testing their performance using FeL‐edge ΔEpeak shifts designed to simulate material defects. These results show that 3D‐CVAE achieves superior anomaly detection and maintains consistent performance across various shift magnitudes. The method demonstrates clear bimodal separation between bulk and anomalous spectra, enabling reliable classification. Further analysis verifies that lower‐dimensional representations are robust to anomalies in the data. While performance advantages over PCA diminish with decreasing anomaly concentration, our method maintains high reconstruction quality even in challenging, noise‐dominated spectral regions. This approach provides a robust framework for unsupervised automated detection of spectral anomalies in EELS‐SI data, particularly valuable for analyzing complex material systems.

Chemistry↗
NASA NTRSDate not supplied

2022 Spring Internship Exit Presentation

As efforts of the National Aeronautics and Space Administration (NASA) and the Federal Aviation Administration (FAA) continue to digitize the air traffic management (ATM) domain, there is countless times of need for downstream natural language processing (NLP) tasks such as named entity recognition, text summarization, classification, and more. Although there are a plethora of open-sourced pre-trained transformer models in the NLP field such as BERT, RoBERTa, XLNet, and GPT-3, these models are trained on general corpora and perform poorly on domain-specific terminology and phraseology seen in ATM documents such as Notice to Airmen (NOTAMs) and Letters of Agreement (LoA). Our proposed research objective will be to first gather a large corpus of air traffic management related documents, orders, notices, books, technical papers, conference papers, articles, and other miscellaneous sources of text data from the FAA, NASA, and accredited conference and publication societies. After gathering this data, many steps will have to be taken to collate and preprocess the data into a format understandable by our test transformer models. Thirdly, we will set up training pipelines to train the RoBERTa model on its unsupervised training task masked language modelling (MLM) using resources provided by the NASA Advanced Supercomputing (NAS) facilities. Finally, these fine-tuned transformer models will be evaluated on their performance on down-stream NLP tasks as mentioned above, to show whether they will be effective when working with ATM related data or not. Once complete, this model could be made open-sourced on the HuggingFace website, where the rest of the ATM community can access and utilize this tool.

NLP↗
DOE OSTI2026

Graph theory inspired anomaly detection at the LHC

Designing model-independent anomaly detection algorithms for analyzing LHC data remains a central challenge in the search for new physics, due to the high dimensionality of collider events. In this work, we develop a graph autoencoder as an unsupervised, model-agnostic tool for anomaly detection, using the LHC Olympics dataset as a benchmark. By representing jet constituents as a graph, we introduce a method to systematically control the information available to the model through sparse graph constructions that serve as physically motivated inductive biases. Specifically, (1) we construct graph autoencoders based on locally rigid Laman graphs and globally rigid unique graphs, and (2) we explore the clustering of jet constituents into subjets to interpolate between high- and low-level input representations. We obtain the best performance, measured in terms of the Significance Improvement Characteristic curve for an intermediate level of subjet clustering and certain sparse unique graph constructions. We further investigate the role of graph connectivity in jet classification tasks. Our results demonstrate the potential of leveraging graph-theoretic insights to refine and increase the interpretability of machine learning tools for collider experiments.

Automation↗
NASA NTRSDate not supplied

Utilization of Machine Learning Techniques for Managing the Tracking and Data Relay Satellite Constellation

National Aeronautics and Space Administration’s (NASA) Goddard Space Flight Center (GSFC) operates a constellation of ten geosynchronous Tracking and Data Relay Satellites (TDRS). The TDRS constellation consists of multiple geosynchronous communication relay satellites located around the equator so they can provide continual coverage of any mission in low earth orbit. The TDRS are located primarily in three oceanic regions around the earth. NASA’s White Sands Complex provides the ground communication support for TDRS located over the Atlantic and Pacific Oceans. Another TDRS ground station in Guam supports the TDRS over the Indian Ocean. With these satellites the TDRS network can provide continuous coverage of satellites in low-earth orbit. The NASA Space Network (SN) project office at GSFC manages the constellation of spacecraft. Major customers of the TDRS constellation include, but are not limited to, the International Space Station and the Hubble Space Telescope. The TDRS constellation has three generations of satellites and has been active for over 30 years providing reliable communication links between customer satellites and corresponding ground stations. However, one of the major concerns for TDRS, and in any space mission, is to ensure the health and safety of the spacecraft. Generally, engineers use telemetry data to monitor and analyze the performance and state of health of the spacecraft. Telemetry data contains hundreds of parameters that monitor each important component in the spacecraft, which can be utilized to recognize and characterize the behavior of the spacecraft. Each parameter contains considerable information to represent time-dependent properties of each spacecraft subsystem and component. During the entire life of a TDRS spacecraft, thousands of gigabytes of telemetry data are transmitted in real-time from the spacecraft to the ground station at the White Sands Complex in Las Cruces, New Mexico, and recorded as historical data sets for engineers to process and analyze the events that occurred on-orbit. These parameters contain the function of multiple spacecraft subsystems, such as the attitude control system (ACS), Thermal, Electrical Power Subsystem (EPS), etc. . The first and second generations have exceeded their required lifetime and NASA is keen to manage these spacecrafts carefully in order to maximize the remaining life using the spacecraft telemetry. The challenge is to know when the risk of losing a spacecraft in geosynchronous orbit exceeds the benefit of continued operations for customer support. In the TDRS fleet, the EPS is the most critical subsystem related to spacecraft operations. Failure of the EPS would strand a spacecraft in geosynchronous orbit. Since EPS provides power to the spacecraft, component failures ultimately lead to the inability to support the spacecraft loads and the communications payload. For instance, TDRS-8 has several anomalies in EPS including the Bus Voltage Limiter (BVL) shunt current, solar array loss of circuits, and failed battery cells. Any of these anomalies can cause critical issues to the spacecraft. Therefore, developing a system to analyze and perform early detection of a potential anomaly is an important issue in telemetry data analysis. In recent years, Telemetry Mining (TM) has been proposed to process telemetry data by using Data Mining (DM) techniques such as classification, clustering, regression and anomaly detection. Anomaly detection, also known as outlier detection, has been widely used in many data mining areas such as remote sensing, medical data processing and digital image processing. The goal of anomaly detection is to detect abnormal data, which contains a relatively low probability of occurrence among the entire data set. Early detection of anomalies is one of the most significant issues in managing the spacecraft configuration. If anomalies can be detected early enough, then the redundant resources can be used to extend the life of the operational spacecraft. We present an unsupervised anomaly detection method to process the EPS data extracted from TDRS-8. This is different from traditional analytical methods, which use telemetry data to illustrate behavior and physical meaning of each spacecraft component. TM connects multiple parameters as a vector and then conducts data analysis on this high dimension telemetry vector. This method is looking at the properties of a high dimensional vector that is able to consider the relationship between different parameters in the anomaly detection problem. This kind of method performs much better than the traditional limit checking method. In addition, we propose a new approach of real-time anomaly detection to process telemetry data in real-time, which can then be applied to spacecraft monitoring with high reliability, low cost and high accuracy.

Machine Learning (ML)↗
DOE OSTI2024

Unpaired image translation to mitigate domain shift in liquid argon time projection chamber detector responses

Deep learning algorithms often are developed and trained on a training dataset and deployed on test datasets. Any systematic difference between the training and a test dataset may severely degrade the final algorithm performance on the test dataset—what is known as the domain shift problem . This issue is prevalent in many scientific domains where algorithms are trained on simulated data but applied to real-world datasets. Typically, the domain shift problem is solved through various domain adaptation (DA) methods. However, these methods are often tailored for a specific downstream task, such as classification or semantic segmentation, and may not easily generalize to different tasks. This work explores the feasibility of using an alternative way to solve the domain shift problem that is not specific to any downstream algorithm. The proposed approach relies on modern Unpaired Image-to-Image (UI2I) translation techniques, designed to find translations between different image domains in a fully unsupervised fashion. In this study, the approach is applied to a domain shift problem commonly encountered in Liquid Argon Time Projection Chamber (LArTPC) detector research when seeking a way to translate samples between two differently distributed LArTPC detector datasets deterministically. This translation allows for mapping real-world data into the simulated data domain where the downstream algorithms can be run with much less domain-shift-related performance degradation. Conversely, using the translation from the simulated data to a real-world domain can increase the realism of the simulated dataset and reduce the magnitude of any systematic uncertainties. To evaluate the quality of the translations, we use both pixel-wise metrics and a downstream task to measure the effectiveness of UI2I methods for mitigating the domain shift problem. We adapted several popular UI2I translation algorithms to work on scientific data and demonstrated the viability of these techniques for solving the domain shift problem with LArTPC detector data. To facilitate further development of DA techniques for scientific datasets, the ‘Simple Liquid-Argon Track Samples’ dataset used in this study is also published.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Deep Learning for Spectroscopic X-ray Nano-Imaging Denoising

Synchrotron transmission X-ray microscopy with absorption near edge structure (TXM-XANES) is a powerful tool for investigating the structure and composition of materials at nano- to meso-scales. It is, however, often challenged by high levels of noise that obscure critical details at the single-pixel level. To address this issue, a deep learning-based algorithm is developed for suppressing the image noise, grounded in self-supervised learning principles. In contrast to traditional image denoising methods, this approach successfully enhances the visibility of fine details while significantly reducing the noise in the X-ray images. Through this advancement, the potential of the approach for improving the accuracy and interpretability of the TXM-XANES data is demonstrated, thereby enabling more precise detection of nanoscale phenomena such as inhomogeneous cation redox and metal segregation in battery cathode materials. This technique offers an effective new avenue for harnessing the full potential of synchrotron TXM-XANES imaging, paving the way for a range of exciting new studies in materials science and beyond.

36 MATERIALS SCIENCE↗
DOE OSTI2024

An explainable variational autoencoder model for three-dimensional acoustic emission source localization in hollow cylindrical structures

We introduce an explainable variational autoencoder for three-dimensional (3D) localization of acoustic emission sources in hollow cylindrical structures, with an unsupervised approach. This research capitalizes on multi-arrival waveforms generated by helical path propagation in cylindrical geometries to enable efficient two-receiver localization. By integrating the modal characteristics of Lamb modes under multi-path conditions, we demonstrate that two sets of time-of-arrival differences and peak amplitudes extracted from one receiver can serve as effective localization features. This initial approach identifies four potential source locations, highlighting the feasibility of two-receiver source localization using traditional feature extraction methods. However, direct extraction can be challenging when mode overlaps occur, complicating the localization process. To address this, our work proposes a novel waveform-based method. This method leverages the consistent dispersion characteristics within isotropic materials, where each unique combination of mode arrival times and peak amplitudes constructs a distinct waveform. This distinctiveness overcomes the ambiguities associated with mode overlaps, significantly enhancing the method’s precision and robustness. Our approach adopts a data-driven strategy for waveform-based localization using variational autoencoder (VAE). VAE discerns waveform patterns for localization, while also addressing data uncertainties. The VAE’s encoder and decoder networks capture the localization process and the source’s influence on waveform generation, respectively, guiding latent variables to segregate waveforms by source in the latent space. The design of the learning process focuses on specific localization characteristics to enhance result explainability. Localization predictions are generated by projecting test waveforms, not included in the training set, onto a trained latent space. The prediction is determined using a nearest-neighbor approach based on the closest latent representation of a source. Validation with pencil-lead-break tests on a metallic pipe confirmed our method’s effectiveness, achieving an averaged 3D localization accuracy of 0.84.

Lee, Guan-Wei↗
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 create more accurate transcripts of air traffic management conversations. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement (LOA) which is used as the evaluation case in this paper. These applications can benefit from the use of state-of-the-art Natural Language Processing (NLP) techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of transferring pre-trained large language models to the aviation domain by adapting transformer based models using aviation datasets. This paper utilized two datasets to adapt a ‘Robustly Optimized Bidirectional Encoder Representations from Transformers Approach’ (RoBERTa) model and two down-stream classification tasks to assess its performance. These datasets are all built upon Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. The first two datasets are used for the adaptation of RoBERTa to the aviation domain and were of different sizes to assess the number of documents needed to adapt to the aviation domain. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the unsupervised adaptation. The second dataset is a separate set of LOA documents with two sets of classification labels to be used for evaluation; one at the document level and one at the line level. These down-stream evaluations allowed the measurement of improvement by adapting RoBERTa. The accuracy increased by 4-6% on both tasks and the F1 score on the class of interest increased by 4-8% from the adaptation.

Air Traffic Management↗
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 create more accurate transcripts of air traffic management conversations. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement (LOA) which is used as the evaluation case in this paper. These applications can benefit from the use of state-of-the-art Natural Language Processing (NLP) techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of transferring pre-trained large language models to the aviation domain by adapting transformer based models using aviation datasets. This paper utilized two datasets to adapt a ‘Robustly Optimized Bidirectional Encoder Representations from Transformers Approach’ (RoBERTa) model and two down-stream classification tasks to assess its performance. These datasets are all built upon Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. The first two datasets are used for the adaptation of RoBERTa to the aviation domain and were of different sizes to assess the number of documents needed to adapt to the aviation domain. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the unsupervised adaptation. The second dataset is a separate set of LOA documents with two sets of classification labels to be used for evaluation; one at the document level and one at the line level. These down-stream evaluations allowed the measurement of improvement by adapting RoBERTa. The accuracy increased by 4-6% on both tasks and the F1 score on the class of interest increased by 4-8% from the adaptation.

Air Traffic Management↗
DOE OSTI2026

Unsupervised anomaly detection in MeV ultrafast electron diffraction

MeV ultrafast electron diffraction (MUED) is a pump-probe technique used to study the dynamic structural evolution of materials. An ultrashort laser pulse triggers structural changes, which are then probed by an ultrashort relativistic electron beam. To overcome low signal-to-noise ratios, diffraction patterns are averaged over thousands of shots. However, shot-to-shot instabilities in the electron beam can distort individual patterns, introducing uncertainty. Improving MUED accuracy requires detecting and removing these anomalous patterns from large datasets. In this work, we developed a fully unsupervised methodology for the detection of anomalous diffraction patterns. Using a convolutional autoencoder, we calculate the reconstruction mean squared error of the diffraction patterns. Based on the statistical analysis of this error, we provide the user an estimation of the probability that the pattern is normal, which also allows a posterior visual inspection of the images that are difficult to classify. This method has been trained with only 100 diffraction patterns and tested on 1521 patterns, resulting in a false positive rate between 0.2% and 0.4%, with a training time of 10 s per image and a test time of about 1 s per image. Here, the proposed methodology can also be applied to other diffraction techniques in which large datasets are collected that include faulty images due to instrumental instabilities.

43 PARTICLE ACCELERATORS↗