Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “data consistent”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Terrestrial laser scanning data (Levels 0 and 1) for Pasoh, Malaysia, Sep 2024

This data package contains data from terrestrial laser scanning (TLS) at the Pasoh Forest Reserve, Malaysia. The Pasoh Forest Reserve is a facility of the Forest Research Institute Malaysia, and contains evergreen lowland dipterocarp forest. The Next-Generation Ecosystem Experiments Tropics (NGEE-Tropics) study areas at Pasoh were established to study how different species respond to climatic variation and soil water availability. Two study areas were chosen representing different topography and species. The TLS data archived here were collected to provide detailed, three-dimensional information about forest structure. Specifically, data were collected to allow tree-level characterization of woody structure and leaf area for 12 focal trees with FloraPulse and sap flux sensors, facilitating estimation of woody biomass and leaf area to allow upscaling of water content and transpiration data to the tree-level. Scan positions were not selected to provide consistent data for non-focal trees with the study areas. This data package contains the following data: - High-level files document further details of the campaign and data package: 1_CampaignSummary.csv provides details about the campaign and study site, 2_ScanAreasDetail.csv provides details about each separate scan area (groups of scans post-processed into a single point cloud), 3_TerrestrialLidarSensor.csv provides further technical details about the Riegl VZ-400i TLS sensor, TLS_CSV_dd.csv is a CSV Data Dictionary providing information about the fields in CSV files following the ESS-DIVE CSV File Formatting Guidelines Reporting Format, TLS_flmd.csv is a File Level Metadata file providing information about each file in the data package following the ESS-DIVE File Level Metadata Reporting Format, and README.txt is a text file describing the overall project and file structure. - Level 0 data are the raw data (.PROJ folders) as recorded by the Riegl VZ-400i TLS instrument before scan co-registration and post-processing with the Riegl's proprietary RiSCAN PRO software, which requires a license. - Level 1 data contain post-processed, co-registered data from each scan area. The "PointClouds" folder for each scan area contains a .las file with 1 cm resolution point cloud data exported from RiSCAN PRO. These are the main files likely to be of interest to most users and can be further processed with any software capable of manipulating .las files (e.g. Python, R CloudCompare). The "Project Information" folder contains log files from post-processing in RiSCAN PRO that may be of interest to users who want to see detailed records of post-processing, including all PDF reports generated by RiSCAN PRO. The "ScanPositions" folder contains information about the final position of all TLS scans, after post-processing, in multiple formats. The file ScanPositions_*.csv provides final geo-referenced scan positions, and the file SOP_backup_*.csv can be used in RiSCAN PRO to restore the co-registered scan positions if users wish to re-process raw data (Level 0 .PROJ folders) with RiSCAN PRO software (e.g., subsample to a different resolution, exclude a certain scan position, or apply different filters on reflectance or deviation values) without redoing time-consuming co-registration steps.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Opportunities for Earth Observation to Inform Risk Management for Ocean Tipping Points

Abstract As climate change continues, the likelihood of passing critical thresholds or tipping points increases. Hence, there is a need to advance the science for detecting such thresholds. In this paper, we assess the needs and opportunities for Earth Observation (EO, here understood to refer to satellite observations) to inform society in responding to the risks associated with ten potential large-scale ocean tipping elements: Atlantic Meridional Overturning Circulation; Atlantic Subpolar Gyre; Beaufort Gyre; Arctic halocline; Kuroshio Large Meander; deoxygenation; phytoplankton; zooplankton; higher level ecosystems (including fisheries); and marine biodiversity. We review current scientific understanding and identify specific EO and related modelling needs for each of these tipping elements. We draw out some generic points that apply across several of the elements. These common points include the importance of maintaining long-term, consistent time series; the need to combine EO data consistently with in situ data types (including subsurface), for example through data assimilation; and the need to reduce or work with current mismatches in resolution (in both directions) between climate models and EO datasets. Our analysis shows that developing EO, modelling and prediction systems together, with understanding of the strengths and limitations of each, provides many promising paths towards monitoring and early warning systems for tipping, and towards the development of the next generation of climate models.

Wood, Richard A. (ORCID:0000000239609513)↗
DOE OSTI2022

Differentially Private K -Means Clustering Applied to Meter Data Analysis and Synthesis

The proliferation of smart meters has resulted in a large amount of data being generated. It is increasingly apparent that methods are required for allowing a variety of stakeholders to leverage the data in a manner that preserves the privacy of the consumers. The sector is scrambling to define policies, such as the so called ‘15/15 rule’, to respond to the need. However, the current policies fail to adequately guarantee privacy. Here, in this paper, we address the problem of allowing third parties to apply K-means clustering, obtaining customer labels and centroids for a set of load time series by applying the framework of differential privacy. We leverage the method to design an algorithm that generates differentially private synthetic load data consistent with the labeled data. We test our algorithm’s utility by answering summary statistics such as average daily load profiles for a 2-dimensional synthetic dataset and a real-world power load dataset.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

DOE EV Data Collection - Vehicle Data

Vehicle data consist of electric vehicle performance data collected directly from the vehicle during standard operations. Data were collected using onboard data loggers that were either installed by the project team or preinstalled by the original equipment manufacturer. Data recorded by the data loggers were made accessible via an online web portal or an application programming interface. Different data loggers were used (HEM, ViriCiti, and Geotab), and the method for each vehicle is defined in the vehicle attributes file. Some systems collected data on a “trip-level” basis, in which each row of a table represents a single trip (the period between a key-on and key-off event), whereas other data were collected on a per-day basis, in which each row represents a single day of operation. Data were collected over a range of data collection periods, depending on the project. Data have been anonymized by removing information or decreasing information resolution as necessary so that fleets are not identifiable. Due to the wide range of vehicle types represented and variation in data collection, data parameters and frequencies differ between vehicles and fleets The **Performance Data Daily/Trip Data Dictionaries** contain definitions for each available parameter associated with a vehicle’s operations, aggregated at either a daily or trip level. The parameters available will vary from vehicle to vehicle, but every possible parameter will be defined. The **Vehicle Attributes Data Dictionary** contains definitions for each available parameter associated with a vehicle’s physical and functional attributes and fleet context. The **Vehicle Attributes** table contains specific vehicle characteristics, coded to an anonymous Vehicle ID. This Vehicle ID can be used as a key between vehicle data and vehicle attribute tables. The **Vehicle Data** tables contain the data from each vehicle’s operations, aggregated at either a daily or trip level, coded to an anonymous Vehicle ID. This Vehicle ID can be used as a key between vehicle data and vehicle attribute tables. Data is being uploaded quarterly through 2023 and subject to change until the conclusion of the project.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Machine learning models of intermittent operation of RO wellhead water treatment for salinity reduction and nitrate removal

Machine learning models were developed for intermittent multi-mode operation of a wellhead reverse osmosis water purification and desalination system to predict salt passage, nitrate passage, and permeate flux. The models, based on long short-term memory (LSTM) recurrent neural network (RNN) architecture, included an attention mechanism to increase model performance in proximity of the regulatory limit for nitrate. Training and testing of the models for the Startup, Production, Shutdown and Flushing operational modes were based on operational data (consisting of 22 process variables per data sample) acquired every 2–5 s over a six-month period. The significant sets of model input attributes for the different operational modes were assessed via Spearman ranking correlation, Self-Organizing Map (SOM) analysis and feed forward feature selection (FFFS). Although the variability of nitrate passage, salt passage and permeate flux was significant over the four operational modes, prediction performance for the three outcomes were with R2 and Average Absolute Relative Error (AARE) of 0.78–0.95 and 2.96–6.16 %, respectively. Model updates post membrane elements replacement demonstrated similar levels of prediction accuracy. The study results suggest that there is merit in exploring the utility of multi-mode models for sensor fault detection, data imputation, and for potential use in model-predictive control.

Intermittent RO operation↗
DOE OSTI2022

Tsunami Early Warning From Global Navigation Satellite System Data Using Convolutional Neural Networks

Abstract We investigate the potential of using Global Navigation Satellite System (GNSS) observations to directly forecast full tsunami waveforms in real time. We train convolutional neural networks to use less than 9 min of GNSS data to forecast the full tsunami waveforms over 6 hr at select locations, and obtain accurate forecasts on a test data set. Our training and test data consists of synthetic earthquakes and associated GNSS data generated for the Cascadia Subduction Zone using the MudPy software, and corresponding tsunami waveforms in Puget Sound computed using GeoClaw. We use the same suite of synthetic earthquakes and waveforms as in earlier work where tsunami waveforms were used for forecasting, and provide a comparison. We also explore varying the number of GNSS stations, their locations, and their observation durations.

Rim, Donsub↗
DOE OSTI2020

Radionuclide Screening Analysis and Transport Parameters for Pahute Mesa Detonations, Nevada National Security Site

A “screening analysis” is implemented at Pahute Mesa (PM), Nevada National Security Site (NNSS), to determine subsets of the 43 radionuclides listed in the radionuclide inventory of Finnegan et al. (2016) that are of relevance or potential relevance to the hydrologic source term (HST) for assessment of radionuclide transport in groundwater. Consideration is added to how levels of contamination are defined and whether any other radionuclides not in the inventory are relevant to the HST. A model is developed to estimate a range of possible source concentrations in groundwater that account for uncertainty in partitioning into melt glass and sorption into surrounding rock of the exchange zone for the 82 PM underground nuclear tests detonated in vertical shafts. The model also accounts for the varied hydrogeochemical settings. Transport parameters needed for the screening model calculations are developed from databases for hydrogeologic units, chemistry, mineralogy, fracture spacing, fracture aperture, fracture openness, matrix porosity, bulk density, saturation, and alteration. Of key importance is consideration of diagenetic zonation of the mineralogy related to hydrothermal alteration. The screening analysis compares model results with available groundwater radiochemistry data. Consistency between screening model results and available data helps validate the model for application to all 43 radionuclides, most of which have no measurements or only non-detect measurements of concentration in groundwater. The screening analysis determines that ten radionuclides – tritium, Sr-90, I-129, Cs-137, U-232, U-233, U-234, U-238, and the total of Pu-239 and Pu-240 are relevant to the HST. Determination of relevance is primarily based on data and/or model results indicating source concentration exceeding a maximum contaminant level (MCL) out to 100 years since the time of detonation. U-238 is relevant to assessment of contamination relative to MCLs for uranium and gross alpha particle activity. Five other radionuclides – C-14, Cl-36, Tc-99, Np-237, and Pu-238 – are determined potentially relevant to the HST based on a 0.1 MCL threshold. The screening analysis recommends additional attention to the natural daughter radionuclides of the uranium series (U-238) and thorium series (Th-232) decay chains, particularly Ra-226 and Ra-228, that were not included the inventory.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2025

PV backsheets survey protocol: A framework for geo-spatial field surveys for bulk material characterization and reliability analysis applied across 41 PV systems

As widespread adoption of photovoltaic (PV) technologies continues, understanding the lifetime of modules is paramount to the viability of the industry as an environmentally conscious alternative to traditional energy generation. Although power degradation can affect the total energy production of a module over its lifetime, module safety failures necessitate the removal of a module leading to a loss of not only the particular asset, but the earning potential of the device. Therefore, it is critical to ensure that the components that provide essential safety functions for PV module operate for their entire rated lifetime. PV backsheets provide necessary electrical insulation to the completed device and failure of this component is cause for a immediate removal of the module. Degradation of the PV module backsheet has led to module safety failures in large-scale installations, costing millions of dollars in damages and lost potential revenue. The spatio-temporal degradation of fielded PV modules is important to study in order to identify which modules within installations are experiencing the greatest exposure conditions and in turn have the highest chance of failure. This paper describes a comprehensive field survey protocol developed for monitoring PV module backsheet performance using solely non-destructive methods in commercial PV fields. The protocol establishes a field naming convention, sampling method, data handling requirements, and measurement procedures. By ensuring consistent data collection practices, the field survey protocol enables research groups to obtain data of uniform quality on backsheet performance over multiple years and locations. In this study, the developed protocol was implemented at forty-one PV sites. Eight different types of airside layer backsheet materials including poly(vinylidene fluoride) (PVDF), acrylic PVDF, poly(tetrafluoroethylene-co-hexafluoropropylene-co-vinylidene fluoride) (THV), poly(vinyl fluoride) (PVF), poly(ethylene terephthalate) (PET), fluoroethylene vinyl ether (FEVE), polyethylene naphthalate (PEN), and glass were identified using attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy. The field survey results show that the spatial distribution of degradation indicators are non-uniform within a particular module, individual site, and across site locations. The degradation of PV modules increased in severity for modules mounted at the edge of rows (across a field) and near the junction box (within a module). This study demonstrates the sensitivity of material performance to exposure length across different materials and climates.

14 SOLAR ENERGY↗
DOE OSTI2021

Search for topological defect dark matter with a global network of optical magnetometers

Ultralight bosons such as axion-like particles are viable candidates for dark matter. They can form stable, macroscopic field configurations in the form of topological defects that could concentrate the dark matter density into many distinct, compact spatial regions that are small compared with the Galaxy but much larger than the Earth. Here we report the results of the search for transient signals from the domain walls of axion-like particles by using the global network of optical magnetometers for exotic (GNOME) physics searches. We search the data, consisting of correlated measurements from optical atomic magnetometers located in laboratories all over the world, for patterns of signals propagating through the network consistent with domain walls. The analysis of these data from a continuous month-long operation of GNOME finds no statistically significant signals, thus placing experimental constraints on such dark matter scenarios.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Experimental Study for Validation of Resilient Response of Controls - Thrust 2 Controls Assessment

The focus of this project is to design, execute, and analyze experiments around resilience response. As part of the Resilience through Data-Driven, Intelligently Designed Control (RD2C) initiative, this project builds upon the previously designed control algorithms and experiments. It designs detailed uses cases to test RD2C-developed mitigation schemes in specific realistic scenarios, analyze and save the measurement data sets. Having a high-confidence and consistent data set is crucial for identifying relevant metrics to describe the system resilience in specific real-life-like scenarios of power grid operation.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

Status of DUNE Offline Computing

We summarize the status of Deep Underground Neutrino Experiment (DUNE) Offline Software and Computing program. We describe plans for the computing infrastructure needed to acquire, catalog, reconstruct, simulate and analyze the data from the DUNE experiment and its prototypes in pursuit of the experiment's physics goals of precision measurements of neutrino oscillation parameters, detection of astrophysical neutrinos, measurement of neutrino interaction properties and searches for physics beyond the Standard Model. In contrast to traditional HEP computational problems, DUNE's Liquid Argon Time Projection Chamber data consist of simple but very large (many GB) data objects which share many characteristics with astrophysical images. We have successfully reconstructed and simulated data from 4% prototype detector runs at CERN. The data volume from the full DUNE detector, when it starts commissioning late in this decade will present memory management challenges in conventional processing but significant opportunities to use advances in machine learning and pattern recognition as a frontier user of High Performance Computing facilities capable of massively parallel processing. Our goal is to develop infrastructure resources that are flexible and accessible enough to support creative software solutions as HEP computing evolves.

43 PARTICLE ACCELERATORS↗
DOE OSTI2026

COMPASS-FME Terrestrial Ecosystem Manipulation to Probe the Effects of Storm Treatments (TEMPEST) Experiment Level 2 Sensor Data v2-1

This is the version v2-1 Level 2 (L2) data release for COMPASS-FME environmental sensors located at our Terrestrial Ecosystem Manipulation to Probe the Effects of Storm Treatments (TEMPEST) experimental site. This manipulative, ecosystem-scale TEMPEST experiment addresses the potential for freshwater and estuarine-water disturbance events to alter tree function, species composition, and ecosystem processes in a deciduous coastal forest in MD, USA. The experiment uses a large-unit (2000 m2), un-replicated experimental design, with three 50 m × 40 m plots serving as control, freshwater, and estuarine-water treatments. Level 2 (L2) data consist of sensor observations from the COMPASS-FME synoptic sites, TEMPEST, and DELUGE. Compared to the L1 data, these are more consistent (always 15-minute timestamps for the entire year); better QA/QC’d (out of bounds, out of service, and extreme outlier values are removed); and more complete, with a gap-filled time series available alongside the main observations, and additional derived (calculated) variables. L2 data are intended to be rapidly and easily usable in analyses and simulations. However, algorithmic outlier identification always carries the risk of removing valid data, and Level 1 data may be more suitable for analyses that focus on variability or extreme events. This dataset includes: - An overall dataset README file that describes the current version, gives citation and contact information, etc. - Site- and year-specific folders, each holding variable-specific Parquet (a high performance, space efficient format; see https://parquet.apache.org) data files for each site and plot in that year. - Metadata files within each site-year folder provide full information on data units, expected ranges, contact information, detailed flood times, as well as a general description of the site. - Environmental sensor types that appear in the data files include weather (ClimaVUE50, CS, RM Young, and LI instruments in the graphs below); soil conditions (TEROS12); soil redox state (Redox); groundwater variables (AquaTROLL200 and AquaTROLL600); open water sondes (Exo); tree sap velocity (Sapflow); and system voltage and state (Datalogger). Data are reported every 15 minutes. Please see v2-1 TEMPEST L2 Sensor Package Quick Start.pdf for detailed information on data package structure, temporal coverage, and versioning. Data files are in Apache Parquet, a high performance, space efficient format for tabular data. These files can be read using R's `arrow` package (https://arrow.apache.org/docs/r/), with similar tools available in other languages. The TEMPEST flood events occurred on the following dates. They lasted for ~10 hours each day and delivered ~80,000 gallons to each plot; many data streams are available at 1 or 5 minute frequency during these periods. * Tests: Aug 25 (fresh plot) and Sep 9 (salt plot), 2021 * TEMPEST 1: June 22, 2022 * TEMPEST 2: June 6-7, 2023 * TEMPEST 3: June 11-13, 2024

EARTH SCIENCE > ATMOSPHERE > ATMOSPHERIC TEMPERATU↗
DOE OSTI2026

COMPASS-FME Synoptic Sites Level 2 Sensor Data v2-1

This is the version 2-1 Level 2 (L2) data release for COMPASS-FME environmental sensors located at our synoptic field sites. COMPASS-FME is studying sites in two distinct regions, the Chesapeake Bay and the Western Lake Erie Basin. We established the network at seven "synoptic" (observational) sites along the Chesapeake Bay and Lake Erie coastlines, collectively generating over three million observations per month, to track and comprehend environmental changes where land and water intersect. Additionally, the two regions provide an interesting contrast of saltwater and freshwater coasts that allow us to differentiate the impacts of inundation and coastal water chemistries in two nationally important coastal systems. Level 2 (L2) data consist of sensor observations from the COMPASS-FME synoptic sites, TEMPEST, and DELUGE. Compared to the L1 data, these are more consistent (always 15-minute timestamps for the entire year); better QA/QC’d (out of bounds, out of service, and extreme outlier values are removed); and more complete, with a gap-filled time series available alongside the main observations, and additional derived (calculated) variables. L2 data are intended to be rapidly and easily usable in analyses and simulations. However, algorithmic outlier identification always carries the risk of removing valid data, and Level 1 data may be more suitable for analyses that focus on variability or extreme events. This dataset includes: - An overall dataset README file that describes the current version, gives citation and contact information, etc. - Site- and year-specific folders, each holding variable-specific Parquet (a high performance, space efficient format; see https://parquet.apache.org) data files for each site and plot in that year. - Metadata files within each site-year folder provide full information on data units, expected ranges, contact information, detailed flood times, as well as a general description of the site. - Environmental sensor types that appear in the data files include weather (ClimaVUE50, CS, RM Young, and LI instruments in the graphs below); soil conditions (TEROS12); soil redox state (Redox); groundwater variables (AquaTROLL200 and AquaTROLL600); open water sondes (Exo); tree sap velocity (Sapflow); and system voltage and state (Datalogger). Data are reported every 15 minutes. Data files are in Apache Parquet, a high performance, space efficient format for tabular data. These files can be read using R's `arrow` package (https://arrow.apache.org/docs/r/), with similar tools available in other languages. Please see v2-1 L2 Sensor Package QStart.pdf for detailed information on data package structure, temporal coverage, and versioning.

EARTH SCIENCE > ATMOSPHERE > ATMOSPHERIC TEMPERATU↗
DOE OSTI2023

DOLCE: A Model-Based Probabilistic Diffusion Framework for Limited-Angle CT Reconstruction

Limited-Angle Computed Tomography (LACT) is a non-destructive evaluation technique used in a variety of applications ranging from security to medicine. The limited angle coverage in LACT is often a dominant source of severe artifacts in the reconstructed images, making it a challenging inverse problem. We present DOLCE, a new deep model-based framework for LACT that uses a conditional diffusion model as an image prior. Diffusion models are a recent class of deep generative models that are relatively easy to train due to their implementation as image denoisers. DOLCE can form high-quality images from severely under-sampled data by integrating data-consistency updates with the sampling updates of a diffusion model, which is conditioned on the transformed limited-angle data. We show through extensive experimentation on several challenging real LACT datasets that, the same pre-trained DOLCE model achieves the SOTA performance on drastically different types of images. Additionally, we show that, unlike standard LACT reconstruction methods, DOLCE naturally enables the quantification of the reconstruction uncertainty by generating multiple samples consistent with the measured data.

Kim, Hyojin↗
DOE OSTI2024

Climate-invariant machine learning

Projecting climate change is a generalization problem: We extrapolate the recent past using physical models across past, present, and future climates. Current climate models require representations of processes that occur at scales smaller than model grid size, which have been the main source of model projection uncertainty. Recent machine learning (ML) algorithms hold promise to improve such process representations but tend to extrapolate poorly to climate regimes that they were not trained on. To get the best of the physical and statistical worlds, we propose a framework, termed “climate-invariant” ML, incorporating knowledge of climate processes into ML algorithms, and show that it can maintain high offline accuracy across a wide range of climate conditions and configurations in three distinct atmospheric models. Our results suggest that explicitly incorporating physical knowledge into data-driven models of Earth system processes can improve their consistency, data efficiency, and generalizability across climate regimes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Data Fusion to Enhance Quality Control and Analysis with Instruments at the Marine and Coastal Research Laboratory

Deploying environmental monitoring instruments in the marine environment can be challenging, facing challenges around device survivability, biofouling and corrosion, and consistent data collection. This project explores the use of data fusion – the process of integrating multiple data sources to produce more consistent, accurate, and useful information – to build a consistent long-term monitoring system at the Marine and Coastal Research Laboratory (MCRL) in Sequim, Washington. Unused instruments that had been acquired from past projects were inventoried and deployments planned on the MCRL pier and floating dock. A total of 8 instruments were deployed including a tide gauge, hydrophone, acoustic Doppler current profiler (ADCP), photosynthetically active radiation (PAR) sensors, meteorological station, and three water quality sensors. Deployments were planned to be well-protected around the pier structure and a maintenance schedule was created for cleaning and recalibration. An automated data pipeline was created to aggregate data on edge computers that push data to Amazon Web Services (AWS) cloud storage every 15 minutes, performing automated quality control and data transformations using the Time Series Data Analytical Toolkit (TSDAT). Continued efforts are underway to maintain this system into the future, take a data-driven approach to maintenance scheduling, improve the reliability of the system, and share the data with a variety of end-users.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Machine learning for seismic low-frequency extrapolation

The cycle-skipping problem that plagues full waveform inversion (FWI) can be at least partially mitigated if low frequencies (which encode the kinematics of wave propagation in seismic data) are recorded. However, seismic sources and receivers are band-limited, so seismic data does not generally include signals down to 0 Hz. To improve our ability to solve the seismic inverse problem, one can synthesize this missing low-frequency (LF) content from the recorded high-frequency (HF) data using machine learning (ML) models. Deep learning models such as convolutional neural networks (CNNs) demonstrate impressive ability to perform low frequency extrapolation. However, such models require powerful hardware (GPU machines) and careful training. We assess the extrapolation capabilities of three different ML models that do not require GPU machines, namely, random forest, Gaussian process regression and gradient boosting, on both synthetic and real data. Experimental results on two synthetic data sets (generated from a low velocity lens embedded in a homogeneous medium, and the Marmousi model) demonstrate that FWI applied to the extrapolated data consistently improves inversion accuracy relative to FWI applied to the original data sets that do not contain low frequencies. Application of low-frequency extrapolation to real data from the Northwest Shelf of Australia demonstrates that tree-based ML models such as gradient boosting can outperform CNNs in terms of both accuracy and computational cost on non-GPU architectures.

58 GEOSCIENCES↗
DOE OSTI2024

Improving Text Classification with Large Language Model-Based Data Augmentation

Large Language Models (LLMs) such as ChatGPT possess advanced capabilities in understanding and generating text. These capabilities enable ChatGPT to create text based on specific instructions, which can serve as augmented data for text classification tasks. Previous studies have approached data augmentation (DA) by either rewriting the existing dataset with ChatGPT or generating entirely new data from scratch. However, it is unclear which method is better without comparing their effectiveness. This study investigates the application of both methods to two datasets: a general-topic dataset (Reuters news data) and a domain-specific dataset (Mitigation dataset). Our findings indicate that: 1. ChatGPT generated new data consistently enhanced model’s classification results for both datasets. 2. Generating new data generally outperforms rewriting existing data, though crafting the prompts carefully is crucial to extract the most valuable information from ChatGPT, particularly for domain-specific data. 3. The augmentation data size affects the effectiveness of DA; however, we observed a plateau after incorporating 10 samples. 4. Combining the rewritten sample with new generated sample can potentially further improve the model’s performance.

97 MATHEMATICS AND COMPUTING↗