Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Adaptation models”

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 1,207 records · Page 67
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Distributed Solar Adoption in Orlando: A Household-Level Model for Distribution Resource Planning

Potential for rooftop solar in Florida is massive (47% of retail sales, 3rd overall nationally), yet adoption lags (12th nationally). A 2018 Florida Public Service Commission ruling authorizing solar third-party ownership (leasing) has substantially increased attention on distributed solar in the state. The city of Orlando has committed to a 100% clean-energy target by 2050 and deployment of solar and storage are expected to contribute significantly to reaching the goal. Deployment of customer-adopted solar, unlike utility-procured solar, is uncertain, but known to be spatially correlated with demographic factors and existing adoption. We develop a new method to adapt NREL's dGen model in order to represent building-level agents in adoption forecasts for the Orlando Utility Commission (OUC) service territory. Using the agent-based model we develop projections of solar adoption, subject to scenarios varying future solar costs and valuation, and aggregate adoption predictions by OUC distribution feeder. We find substantial spatial heterogeneity in the projected level of adoption by OUC distribution feeder. For instance, 25% of all projected adoption through 2050 would be concentrated on just 5% of feeders and 88% of projected adoption on 50% of feeders. Because of the uncertainty in adoption, bottoms-up solar adoption forecasting methods at the household-level are integral to long-term resource planning by anticipating system needs as customers increasingly adopt distributed solar, storage, electric vehicles, and other distributed energy resources.

14 SOLAR ENERGY↗
DOE OSTI2024

M3SF-24LL010302052-Comprehensive Analysis of Radionuclide Interaction with Hydrothermally Altered Repository Materials

This progress report (Level 3 Milestone Number M3SF-24LL010302052) summarizes research conducted at Lawrence Livermore National Laboratory (LLNL) within the Crystalline Work Package Number SF-24LL01030205. The research is focused on actinide and radionuclide sequestration in hydrothermally altered repository materials. In FY24, we completed a rigorous analysis of Se sorption to iron oxide phases using our L-SCIE sorption database. This effort explicitly accounts for surface titration behavior of oxide surfaces that was the subject of a recently published manuscript (Han et al., 2023). With this in mind, our Se sorption analysis now yields a more robust workflow for developing self-consistent surface complexation modeling approaches that can be adapted to specific SCM conceptual and numerical approaches (i.e. non-electrostatic, diffuse layer, triple layer models). In late FY24, we plan to publish the results of our comprehensive surface complexation modeling of Se(IV) and Se(VI) sorption to iron oxide mineral phases and provide a path forward to developing robust radionuclide sorption models for use in performance assessment. In FY24, we also submitted a manuscript summarizing our approach to integrating radionuclide sorption and coprecipitation phenomena and evaluation of radionuclide partitioning values across a range of radionuclides relevant to performance assessment. We demonstrated our approach in detail using Se sorption and coprecipitation with iron oxide minerals as a test case. This manuscript was recently accepted for publication in Applied Geochemistry (Balboni et al., Accepted). We also continued experiments to identify radionuclide interaction with hydrothermally altered crystalline repository and backfill materials. Recent research performed at Los Alamos National Laboratory (LANL) and Sandia National Laboratory (SNL) has provided key insights regarding the hydrothermal alteration behavior of bentonite backfill in the presence of repository materials (steel, concrete, etc.). We are now examining how mineral alteration affects retardation behavior of a suite of radionuclides of interest to repository performance assessment. These experiments also allow us to test the predictive ability of our component additivity approach to surface complexation and ion exchange. Our guiding hypothesis is that a robust surface complexation/ion exchange model and associated database, developed using our L-SCIE approach, can effectively predict changes in radionuclide sorption behavior resulting from the hydrothermal alteration of mineralogy in a repository near field. A short update of results to date is presented below.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2021

Shifts in hydroclimatology of US megaregions in response to climate change

Most of the population and economic growth in the United States occurs in megaregions as the clustered metropolitan areas, whereas climate change may amplify negative impacts on water and natural resources. This study assesses shifts in regional hydroclimatology of fourteen US megaregions in response to climate change over the 21st century. Hydroclimatic projections were simulated using the Variable Infiltration Capacity (VIC) model driven by three downscaled climate models from the Multivariate Adaptive Constructed Analogs (MACA) dataset to cover driest to wettest future conditions in the conterminous United States (CONUS). Shifts in the regional hydroclimatolgy and basin characteristics of US megaregions were represented as a combination of changes in the aridity and evaporative indices using the Budyko framework and Fu's equation. Changes in the climate types of US megaregions were estimated using the Fine Gaussian Support Vector Machine (SVM) method. The results indicate that Los Angeles, San Diego, and San Francisco are more likely to experience less arid conditions with some shifts from Continental to Temperate climate type while the hydroclimatology of Houston may become drier with some shifts from Temperate to Continental climate type. Additionally, water yield is likely to decrease in Seattle. Change in the hydroclimatology of Denver and Phoenix highly depends on the selected climate model. However, the basin characteristics of Phoenix have the highest sensitivity to climate change. Overall, the hydroclimatic conditions of Los Angeles, San Diego, Phoenix, Denver, and Houston have the highest sensitivity to climate change. Understanding of future shifts in hydroclimatology of megaregions can help decision-makers to attenuate negative consequences by implementing appropriate adaptation strategies, particularly in the water-scare megaregions.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Multiscale Electricity Modeling for Evaluating Carbon Capture and Sequestration Technologies (MEME-CCS)

This effort employs and adapts an existing, rigorous multiscale electricity modeling platform at the National Renewable Energy Laboratory (NREL) to evaluate carbon capture and sequestration (CCS) and negative emissions technologies (NET) from the ARPA-E FLECCS program. NREL's modeling platform includes the Regional Energy Deployment System (ReEDS) electric sector capacity expansion model, which projects future electricity generation mixes at sub-state-level resolution that are downscaled to the unit-level to enable hourly, zonal or nodal electricity production cost modeling in the PLEXOS model. The resulting hourly price data from PLEXOS is provided to technology developers under the ARPA-E FLECCS program to enable technology-specific economic analysis. The ReEDS-PLEXOS modeling suite is well-established for examining electric sector futures with high renewable energy penetrations, energy storage, electrification, and distributed generation. The key advancement proposed herein utilizes collaboration with CCS experts at the University of Wyoming and the FLECCS teams to create innovative methods for representing CCS and NET in the ReEDS and PLEXOS models. Expanded technology options, new operational parameterizations, and detailed data defining CO2 capture, transportation, and storage systems are being integrated into these models to allow an unprecedented combination of scope and resolution for exploring the future of CCS and NET. The resulting capabilities will take advantage of high-performance computing resources to permit wide-ranging scenario analysis to assess CCS and NET deployment under alternative CO2 prices, fossil fuel prices, electricity demand growth, and other electric sector characteristics. Final outcomes will include publicly available hourly grid operation and price data for any U.S. region of interest along with open-access capacity expansion tools for evaluating CCS/NET systems. These products will help an emerging CCS/NET industry in the United States by providing economics-driven guidance to technology developers while informing policy and investment decisions in the public and private sectors.

air capture↗
NASA NTRS1997

Vertical Transport Processes for Inert and Scavenged Species: TRACE-A Measurements

The TRACE-A mission of the NASA DC-8 aircraft made a large-scale survey of the tropical and subtropical atmosphere in September and October of 1992. Both In-situ measurements of CO (G. Sachsen NASA Langley) and aerosol size (J. Browell group, NASA Langley) provide excellent data sets with which to constrain vertical transport by planetary boundary layer mixing and deep-cloud cumulus convection. Lidar profiles of aerosol-induced scattering and ozone (also by Bremen) are somewhat require more subtle interpretation as tracers, but the vertical information on layering largely compensates for these complexities. The reason this DC-8 dataset is so useful is that very large areas of biomass burning over Africa and South America provide surface sources of appropriate sizes with which to characterize vertical and horizontal motions; the major limitation of our source description is that biomass burning patterns move considerably every few days, and daily burning inventories are a matter of concurrent, intensive research. We use the Penn State / NCAR MM5 model in an assimilation mode on the synoptic and intercontinental scale, and assess the success it shows in vertical transport descriptions. We find that the general level of emissions suggested by the climatological approach (Will. Has, U. of Montana) appears to be approximately correct, possibly a bit low, for this October, 1992, time period. Vertical transport in planetary boundary layer mixing to 5.5 kin was observed and reproduced in our simulations. Furthermore we find evidence that Blackader "transilient" or matrix-transport scheme is needed, but may require some adaptation in our tracer model: CO seems to exhibit very high values at the top of the planetary boundary layer, a process that stretches the eddy-diffusion parameterization. We will report on progress in improving the deep convective transport of carbon monoxide: the Grail scheme as we used it at 100 kin resolution did not transport enough material to the upper troposphere. We expect to be able to attribute this to either parameterization reasons (inadequacy of this parameterization at the large 100km scale) or other reasons. Nevertheless, the qualitative nature of deep transport by clouds shows up well in the simulations. As for scavengable species, the simulations predict tens of micrograms per standard cubic meter of smoke aerosol in the boundary layer. In a straightforward illustration of our simple bulk-mass scavenging parameterization, to one or two micrograms per standard cubic meter of smoke aerosol in the free troposphere just above the source regions: very high concentrations for the free troposphere. We expect to report on comparisons of these predictions to a variety of observations.

Chatfield, Robert B.↗
DOE OSTI2021

Local models for scatter estimation and descattering in polyenergetic X-ray tomography

We propose a new modeling approach for scatter estimation and descattering in polyenergetic X-ray computed tomography (CT) based on fitting models to local neighborhoods of a training set. X-ray CT is widely used in medical and industrial applications. X-ray scatter, if not accounted for during reconstruction, creates a loss of contrast in CT reconstructions and introduces severe artifacts including cupping, shading, and streaks. Even when these qualitative artifacts are not apparent, scatter can pose a major obstacle in obtaining quantitatively accurate reconstructions. Our approach to estimating scatter is, first, to generate a training set of 2D radiographs with and without scatter using particle transport simulation software. To estimate scatter for a new radiograph, we adaptively fit a scatter model to a small subset of the training data containing the radiographs most similar to it. We compared local and global (fit on full data sets) versions of several X-ray scatter models, including two from the recent literature, as well as a recent deep learning-based scatter model, in the context of descattering and quantitative density reconstruction of simulated, spherically symmetrical, single-material objects comprising shells of various densities. Our results show that, when applied locally, even simple models provide state-of-the-art descattering, reducing the error in density reconstruction due to scatter by more than half.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2020

Algorithms and Statistical Models for Scientific Discovery in the Petabyte Era

The field of astronomy has arrived at a turning point in terms of size and complexity of both datasets and scientific collaboration. Commensurately, algorithms and statistical models have begun to adapt --- e.g., via the onset of artificial intelligence --- which itself presents new challenges and opportunities for growth. This white paper aims to offer guidance and ideas for how we can evolve our technical and collaborative frameworks to promote efficient algorithmic development and take advantage of opportunities for scientific discovery in the petabyte era. We discuss challenges for discovery in large and complex data sets; challenges and requirements for the next stage of development of statistical methodologies and algorithmic tool sets; how we might change our paradigms of collaboration and education; and the ethical implications of scientists' contributions to widely applicable algorithms and computational modeling. We start with six distinct recommendations that are supported by the commentary following them. This white paper is related to a larger corpus of effort that has taken place within and around the Petabytes to Science Workshops (https://petabytestoscience.github.io/).

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

The Application of a Genetic Algorithm to the Optimization of a Mesoscale Model for Emergency Response

Besides solving the equations of momentum, heat, and moisture transport on the model grid, mesoscale weather models must account for subgrid-scale processes that affect the resolved model variables. These are simulated with model parameterizations, which often rely on values preset by the user. Such “free” model parameters, along with others set to initialize the model, are often poorly constrained, requiring that a user select each from a range of plausible values. Finding the values to optimize any forecasting tool can be accomplished with a search algorithm, and one such process—the genetic algorithm (GA)—has become especially popular. As applied to modeling, GAs represent a Darwinian process: an ensemble of simulations is run with a different set of parameter values for each member, and the members subsequently judged to be most accurate are selected as “parents” who pass their parameters onto a new generation. At the U.S. Department of Energy’s Savannah River Site in South Carolina, we are applying a GA to the Regional Atmospheric Modeling System (RAMS) mesoscale weather model, which supplies input to a model to simulate the dispersion of an airborne contaminant as part of the site’s emergency response preparations. An ensemble of forecasts is run each day, weather data are used to “score” the individual members of the ensemble, and the parameters from the best members are used for the next day’s forecasts. As meteorological conditions change, the parameters change as well, maintaining a model configuration that is best adapted to atmospheric conditions. Significance Statement We wanted to develop a forecasting system by which a weather model is run over the Savannah River Site each day and repeatedly adjusted according to how well it performed the previous day. To run the model, a series of values (parameters) must be set to control how the model will calculate winds, temperatures, and other desired variables. Each day the model was run several times using different combinations of these parameters and later compared with observed meteorological conditions. Parameters that produced the most accurate forecasts were preferentially reused to create the forecasts for the next day. The process was tested for the summer of 2020 and exhibited lower errors than forecasts produced by the model using default values of the parameters.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS1994

Anticipatory Neurofuzzy Control

Technique of feedback control, called "anticipatory neurofuzzy control," developed for use in controlling flexible structures and other dynamic systems for which mathematical models of dynamics poorly known or unknown. Superior ability to act during operation to compensate for, and adapt to, errors in mathematical model of dynamics, changes in dynamics, and noise. Also offers advantage of reduced computing time. Hybrid of two older fuzzy-logic control techniques: standard fuzzy control and predictive fuzzy control.

Mccullough, Claire L.↗
NASA NTRS2003

Fast-Running Aeroelastic Code Based on Unsteady Linearized Aerodynamic Solver Developed

The NASA Glenn Research Center has been developing aeroelastic analyses for turbomachines for use by NASA and industry. An aeroelastic analysis consists of a structural dynamic model, an unsteady aerodynamic model, and a procedure to couple the two models. The structural models are well developed. Hence, most of the development for the aeroelastic analysis of turbomachines has involved adapting and using unsteady aerodynamic models. Two methods are used in developing unsteady aerodynamic analysis procedures for the flutter and forced response of turbomachines: (1) the time domain method and (2) the frequency domain method. Codes based on time domain methods require considerable computational time and, hence, cannot be used during the design process. Frequency domain methods eliminate the time dependence by assuming harmonic motion and, hence, require less computational time. Early frequency domain analyses methods neglected the important physics of steady loading on the analyses for simplicity. A fast-running unsteady aerodynamic code, LINFLUX, which includes steady loading and is based on the frequency domain method, has been modified for flutter and response calculations. LINFLUX, solves unsteady linearized Euler equations for calculating the unsteady aerodynamic forces on the blades, starting from a steady nonlinear aerodynamic solution. First, we obtained a steady aerodynamic solution for a given flow condition using the nonlinear unsteady aerodynamic code TURBO. A blade vibration analysis was done to determine the frequencies and mode shapes of the vibrating blades, and an interface code was used to convert the steady aerodynamic solution to a form required by LINFLUX. A preprocessor was used to interpolate the mode shapes from the structural dynamic mesh onto the computational dynamics mesh. Then, we used LINFLUX to calculate the unsteady aerodynamic forces for a given mode, frequency, and phase angle. A postprocessor read these unsteady pressures and calculated the generalized aerodynamic forces, eigenvalues, and response amplitudes. The eigenvalues determine the flutter frequency and damping. As a test case, the flutter of a helical fan was calculated with LINFLUX and compared with calculations from TURBO-AE, a nonlinear time domain code, and from ASTROP2, a code based on linear unsteady aerodynamics.

Reddy, T. S. R.↗
DOE OSTI2024

Hexagonal Lattices of HIV Capsid Proteins Explored by Simulations Based on a Thermodynamically Consistent Model

HIV capsid proteins (CAs) may self-assemble into a variety of shapes under in vivo and in vitro conditions. Here, we employed simulations based on a residue-level coarse-grained (CG) model with full conformational flexibility to investigate hexagonal lattices, which are the underlying structural pattern for CA aggregations. Facilitated by enhanced sampling simulations to rigorously calculate CA dimerization and polymerization affinities, we calibrated our model to reproduce the experimentally measured affinities. Here using the calibrated model, we performed unbiased simulations on several large systems consisting of 1512 CA subunits, allowing reversible binding and unbinding of the CAs in a thermodynamically consistent manner. In one simulation, a preassembled hexagonal CA sheet developed spontaneous curvatures reminiscent of those observed in experiments, and the edges of the sheet exhibited local curvatures larger than those of the interior. In other simulations starting with randomly distributed CAs at different concentrations, existing CA assemblies grew by binding free capsomeres to the edges and by merging with other assemblies. At high CA concentrations, rapid establishment of predominant aggregates was followed by much slower adjustments toward more regular hexagonal lattices, with increasing numbers of intact CA hexamers and pentamers being formed. Our approach of adapting a general CG model to specific systems by using experimental binding data represents a practical and effective strategy for simulating and elucidating intricate protein aggregations.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

A Comparative Study of Physics‐Informed and Data‐Driven Neural Networks for Compound Flood Simulation at River‐Ocean Interfaces: A Case Study of Hurricane Irene

Simulating compound flooding (CF) at the river-ocean interface within large-scale Earth System Models (ESMs) presents significant challenges due to complex interactions between river discharge, storm surge, and tides. This study assesses the comparative advantages of physics-informed and data-driven machine learning (ML) approaches for enhancing local ESM performance. We systematically compare data-driven neural network models (i.e., CNNs, U-Net, Long Short-Term Memory (LSTM), Gated Recurrent Unit), and physics-informed neural network (PINN) models, including vanilla PINN and a finite-difference-based PINN (FD-PINN). Specifically, FD-PINN is introduced to enhance computational efficiency, accelerating vanilla PINNs by ∼6.5 times while improving accuracy. To enhance data-driven model training, a new data-generation approach is developed to sample historical fluvial and coastal flood events, which ensures a robust data set for extreme event prediction. The models are evaluated using a realistic one-dimensional river domain extracted from an ESM's river mesh and the Hurricane Irene event as an independent test case. Results show that FD-PINN achieves accurate predictions with significantly reduced computational costs relative to vanilla PINNs. Among data-driven models, the best overall performance is achieved by a CNN-LSTM hybrid, which balances accuracy and efficiency. While a fully connected CNN (CNN-FC) provides the best accuracy, it incurs high computational cost. Architectures lacking strong temporal modeling tend to underperform on unseen events. These findings highlight the importance of sequence-aware designs for robust generalization. This study reveals the trade-offs between physics-informed and data-driven models and proposes an adaptive hybrid framework for integrating ML into ESMs to enhance local flood simulations.

Earth Systems Modeling↗
DOE OSTI2025

A neural-network-enhanced parameter-varying framework for multi-objective model predictive control applied to buildings

Management of the electrical grid is becoming more complex due to the increased penetration of alternative energy generation technologies and a broadening diversity of electric loads. This complexity creates challenges in balancing demand and generation that can increase the potential for grid instabilities. One effective way to address this issue is to leverage previously unexploited demand flexibility through advanced control strategies. In this work, we propose an advanced control method, called adaptive neural parameter-varying model predictive control (ANPV-MPC), to control the temperature and energy consumption of a building via its Heating, Ventilation, and Air Conditioning system. ANPV-MPC combines key ideas in parameter-varying control, adaptive control, and online learning strategies to bridge the gap between computationally efficient linear model predictive control and more accurate nonlinear model predictive control. The novelty in ANPV-MPC is the use of a physics-inspired Bayesian neural network to estimate the coefficients of the parameter-varying linear control model. The Bayesian neural network additionally provides uncertainty estimates, triggering online training to capture evolving building system conditions. We show that ANPV-MPC can approximate the building system dynamics with a 28.39% higher accuracy than traditional linear model predictive control, resulting in 36.23% better control performance without increasing complexity of the optimal control problem. ANPV-MPC also adapts in real time to previously unseen conditions using online learning, further improving its performance.

24 POWER TRANSMISSION AND DISTRIBUTION↗
NASA NTRS2003

Active Control of High Frequency Combustion Instability in Aircraft Gas-Turbine Engines

Active control of high-frequency (greater than 500 Hz) combustion instability has been demonstrated in the NASA single-nozzle combustor rig at United Technologies Research Center. The combustor rig emulates an actual engine instability and has many of the complexities of a real engine combustor (i.e. actual fuel nozzle and swirler, dilution cooling, etc.) In order to demonstrate control, a high-frequency fuel valve capable of modulating the fuel flow at up to 1kHz was developed. Characterization of the fuel delivery system was accomplished in a custom dynamic flow rig developed for that purpose. Two instability control methods, one model-based and one based on adaptive phase-shifting, were developed and evaluated against reduced order models and a Sectored-1-dimensional model of the combustor rig. Open-loop fuel modulation testing in the rig demonstrated sufficient fuel modulation authority to proceed with closed-loop testing. During closed-loop testing, both control methods were able to identify the instability from the background noise and were shown to reduce the pressure oscillations at the instability frequency by 30%. This is the first known successful demonstration of high-frequency combustion instability suppression in a realistic aero-engine environment. Future plans are to carry these technologies forward to demonstration on an advanced low-emission combustor.

Corrigan, Bob↗
NASA NTRS2018

A Conservative, Scalable, Space-Time Blade Element Rotor Model for Multi-Rotor Vehicles

The development of a parallel blade-element rotor model and its implementation into an adaptive Cartesian method is described. The unsteady version of the rotor model applies a body force to all cells contained in the swept space-time volume at each timestep and special care is taken to maintain axisymmetry on the Cartesian grid. Mesh convergence of rotor thrust and torque is obtained with around 10000 cells in the disk for the steady model. Parallelization is accomplished using OpenMP and the rotor force computation is distributed across all available nodes. Simulations of an isolated XV-15 rotor in hover show good correlation with experimental data and predictions of multi-rotor thrust variation closely match previous high fidelity simulations. The final paper will also include results from the unsteady rotor model and parallel scaling tests.

Conservative↗
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↗
NASA NTRS1987

Adaptive control and noise suppression by a variable-gain gradient algorithm

An adaptive control system based on normalized LMS filters is investigated. The finite impulse response of the nonparametric controller is adaptively estimated using a given reference model. Specifically, the following issues are addressed: The stability of the closed loop system is analyzed and heuristically established. Next, the adaptation process is studied for piecewise constant plant parameters. It is shown that by introducing a variable-gain in the gradient algorithm, a substantial reduction in the LMS adaptation rate can be achieved. Finally, process noise at the plant output generally causes a biased estimate of the controller. By introducing a noise suppression scheme, this bias can be substantially reduced and the response of the adapted system becomes very close to that of the reference model. Extensive computer simulations validate these and demonstrate assertions that the system can rapidly adapt to random jumps in plant parameters.

Merhav, S. J.↗
NASA NTRS2010

Assimilation of Satellite-Derived Skin Temperature Observations into Land Surface Models

Land surface (or "skin") temperature (LST) lies at the heart of the surface energy balance and is a key variable in weather and climate models. Here we assimilate LST retrievals from the International Satellite Cloud Climatology Project (ISCCP) into the Noah and Catchment (CLSM) land surface models using an ensemble-based, off-line land data assimilation system. LST is described very differently in the two models. A priori scaling and dynamic bias estimation approaches are applied because satellite and model LST typically exhibit different mean values and variability. Performance is measured against 27 months of in situ measurements from the Coordinated Energy and Water Cycle Observations Project at 48 stations. LST estimates from Noah and CLSM without data assimilation ("open loop") are comparable to each other and superior to that of ISCCP retrievals. For LST, RMSE values are 4.9 K (CLSM), 5.6 K (Noah), and 7.6 K (ISCCP), and anomaly correlation coefficients (R) are 0.62 (CLSM), 0.61 (Noah), and 0.52 (ISCCP). Assimilation of ISCCP retrievals provides modest yet statistically significant improvements (over open loop) of up to 0.7 K in RMSE and 0.05 in anomaly R. The skill of surface turbulent flux estimates from the assimilation integrations is essentially identical to the corresponding open loop skill. Noah assimilation estimates of ground heat flux, however, can be significantly worse than open loop estimates. Provided the assimilation system is properly adapted to each land model, the benefits from the assimilation of LST retrievals are comparable for both models.

Reichle, Rolf H.↗