Search NASA⌕ Search

SEARCH · Search NASA

Results for “operator inference”

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 379 records · Page 21

Modeling of particle transport, neutrals and radiation in magnetically-confined plasmas with Aurora

In this work, we present Aurora, an open-source package for particle transport, neutrals and radiation modeling in magnetic confinement fusion plasmas. Aurora's modern multi-language interface enables simulations of 1.5D impurity transport within high-performance computing frameworks, particularly for the inference of particle transport coefficients. A user-friendly Python library allows simple interaction with atomic rates from the Atomic Data and Atomic Structure database as well as other sources. This enables a range of radiation predictions, both for power balance and spectroscopic analysis. We discuss here the superstaging approximation for complex ions, as a way to group charge states and reduce computational cost, demonstrating its wide applicability within the Aurora forward model and beyond. Aurora also facilitates neutral particle analysis, both from experimental spectroscopic data and other simulation codes. Leveraging Aurora's capabilities to interface SOLPS-ITER results, we demonstrate that charge exchange is unlikely to affect the total radiated power from the ITER core during high performance operation. Finally, we describe the ImpRad module in the one modeling framework for integrated task framework, developed to enable experimental analysis and transport inferences on multiple devices using Aurora.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Learning and tuning fuzzy logic controllers through reinforcements

A new method for learning and tuning a fuzzy logic controller based on reinforcements from a dynamic system is presented. In particular, our Generalized Approximate Reasoning-based Intelligent Control (GARIC) architecture: (1) learns and tunes a fuzzy logic controller even when only weak reinforcements, such as a binary failure signal, is available; (2) introduces a new conjunction operator in computing the rule strengths of fuzzy control rules; (3) introduces a new localized mean of maximum (LMOM) method in combining the conclusions of several firing control rules; and (4) learns to produce real-valued control actions. Learning is achieved by integrating fuzzy inference into a feedforward network, which can then adaptively improve performance by using gradient descent methods. We extend the AHC algorithm of Barto, Sutton, and Anderson to include the prior control knowledge of human operators. The GARIC architecture is applied to a cart-pole balancing system and has demonstrated significant improvements in terms of the speed of learning and robustness to changes in the dynamic system's parameters over previous schemes for cart-pole balancing.

Berenji, Hamid R.↗

Analyzing spacecraft configurations through specialization and default reasoning

For an intelligent system to describe a real-world situation using as few statements as possible, it is necessary to make inferences based on observed data and to incorporate general knowledge of the reasoning domain into the description. These reasoning processes must reduce several levels of specific descriptions into only those few that most precisely describe the situation. Moreover, the system must be able to generate descriptions in the absence of data, as instructed by certain rules of inference. The deductions applied by the system, then, generate a high-level description from the low-level evidence provided by the real and default data sources. An implementation of these ideas in a real-world situation is described. The application concerns evaluation of Space Shuttle electromechanical system configurations by console operators in the Mission Control Center. A production system provides the reasoning mechanism through which the default assignments and specializations occur. Examples are provided within this domain for each type of inference, and the suitability is discussed of each toward achieving the goal of describing a situation in the fewest statements possible. Finally, several enhancements are suggested that will further increase the intelligence of similar spacecraft monitoring applications.

Barry, Matthew R.↗

Distributed-Memory Sparse Deep Neural Network Inference Using Global Arrays

Partitioned Global Address Space (PGAS) models exhibit tremendous promise in developing efficient and productive distributed-memory parallel applications. They have been used extensively in scientific computations due to conveniently offering a ``shared-memory''-like model and convenient interfaces that separate communication with synchronization. Traditionally, PGAS communication models have been applied to dense/contiguously distributed data, but most modern applications depict varied levels of sparsity. Existing PGAS models require certain adaptations to support distributed sparse computations, since associated computations often require matrix arithmetic, in addition to data movement. The Global Arrays toolkit from Pacific Northwest National Laboratory (PNNL) is one of the earliest PGAS models to combine one-sided data communication and distributed matrix operations and is still used in the popular NWChem quantum chemistry suite. Recently, we have expanded the Global Arrays toolkit to support common sparse operations, like sparse matrix-dense matrix multiplies (SpMM), sparse matrix-sparse matrix multiplication (SpGEMM) and Sampled Dense-Dense Matrix Multiplication (SDDMM). As it turns out, these operations are the bedrock of sparse Deep Learning (DL); sparse deep neural networks and Graph Neural Networks (GNNs) have gained increasing attention recently in achieving speedups on training and inference with reduced memory footprints. Unlike scientific applications in High Performance Computing (HPC), modern (distributed-memory capable) DL toolkits often rely on non-standardized and closed-source vendor software optimizations, creating challenges in software-hardware co-design at scale. Our goal is to support a variety of distributed-memory sparse matrix operations and helper functions in the newly created Sparse Global Arrays (SGA), such that it is possible to build portable and productive Machine Learning scenarios for algorithm/software and hardware codesign purposes. Contemporary data-parallel schemes for training/inference are undergoing a major overhaul since model replication limits scalability and causes resource inefficiencies. As such, we have adopted tensor parallelism in decomposing the model and inputs, to mitigate memory issues. Current implementation is built on top of MPI and uses CPUs to maximize the portability across the platforms.

Distributed computing, machine learning↗

Learning and tuning fuzzy logic controllers through reinforcements

This paper presents a new method for learning and tuning a fuzzy logic controller based on reinforcements from a dynamic system. In particular, our generalized approximate reasoning-based intelligent control (GARIC) architecture (1) learns and tunes a fuzzy logic controller even when only weak reinforcement, such as a binary failure signal, is available; (2) introduces a new conjunction operator in computing the rule strengths of fuzzy control rules; (3) introduces a new localized mean of maximum (LMOM) method in combining the conclusions of several firing control rules; and (4) learns to produce real-valued control actions. Learning is achieved by integrating fuzzy inference into a feedforward neural network, which can then adaptively improve performance by using gradient descent methods. We extend the AHC algorithm of Barto et al. (1983) to include the prior control knowledge of human operators. The GARIC architecture is applied to a cart-pole balancing system and demonstrates significant improvements in terms of the speed of learning and robustness to changes in the dynamic system's parameters over previous schemes for cart-pole balancing.

Berenji, Hamid R.↗

Projecting Future Energy Production from Operating Wind Farms in North America. Part I: Dynamical Downscaling

Abstract New simulations at 12-km grid spacing with the Weather and Research Forecasting (WRF) Model nested in the MPI Earth System Model (ESM) are used to quantify possible changes in wind power generation potential as a result of global warming. Annual capacity factors (CF; measures of electrical power production) computed by applying a power curve to hourly wind speeds at wind turbine hub height from this simulation are also used to illustrate the pitfalls in seeking to infer changes in wind power generation directly from low-spatial-resolution and time-averaged ESM output. WRF-derived CF are evaluated using observed daily CF from operating wind farms. The spatial correlation coefficient between modeled and observed mean CF is 0.65, and the root-mean-square error is 5.4 percentage points. Output from the MPI-WRF Model chain also captures some of the seasonal variability and the probability distribution of daily CF at operating wind farms. Projections of mean annual CF (CF A ) indicate no change to 2050 in the southern Great Plains and Northeast. Interannual variability of CF A increases in the Midwest, and CF A declines by up to 2 percentage points in the northern Great Plains. The probability of wind droughts (extended periods with anomalously low production) and wind bonus periods (high production) remains unchanged over most of the eastern United States. The probability of wind bonus periods exhibits some evidence of higher values over the Midwest in the 2040s, whereas the converse is true over the northern Great Plains. Significance Statement Wind energy is playing an increasingly important role in low-carbon-emission electricity generation. It is a “weather dependent” renewable energy source, and thus changes in the global atmosphere may cause changes in regional wind power production (PP) potential. We use PP data from operating wind farms to demonstrate that regional simulations exhibit skill in capturing actual power production. Projections to the middle of this century indicate that over most of North America east of the Rocky Mountains annual expected PP is largely unchanged, as is the probability of extended periods of anomalously high or low production. Any small declines in annual PP are of much smaller magnitude than changes due to technological innovation over the last two decades.

Meteorology & Atmospheric Sciences↗

Inferred inflow forecast horizons guiding reservoir release decisions across the United States

Abstract. Medium- to long-range forecasts often guide reservoir release decisions to support water management objectives, including mitigating flood and drought risks. While there is a burgeoning field of science targeted at improving forecast products and associated decision support models, data describing how and when forecasts are applied in practice remain undeveloped. This lack of knowledge may prevent hydrological modelers from developing accurate reservoir release schemes for large-scale, distributed hydrology models that are increasingly used to assess the vulnerabilities of large regions to hydrological stress. We address this issue by estimating seasonally varying, regulated inflow forecast horizons used in the operations of more than 300 dams throughout the conterminous United States (CONUS). For each dam, we take actual forward observed inflows (perfect foresight) as a proxy for forecasted flows available to the operator and then identify for each week of the year the forward horizon that best explains the release decisions taken. Resulting “horizon curves” specify for each dam the inferred inflow forecast horizon as a function of the week of the water year. These curves are analyzed for strength of evidence for contribution of medium- to long-range forecasts in decision making. We use random forest classification to estimate that approximately 80 % of large dams and reservoirs in the US (1553±50 out of 1927 dams with at least 10 Mm3 storage capacity) adopt medium- to long-range inflow forecasts to inform release decisions during at least part of the water year. Long-range forecast horizons (more than 6 weeks ahead) are detected in the operations of reservoirs located in high-elevation regions of the western US, where snowpack information likely guides the release. A simulation exercise conducted on four key western US reservoirs indicates that forecast-informed models of reservoir operations may outperform models that neglect the horizon curve – including during flood and drought conditions.

54 ENVIRONMENTAL SCIENCES↗

Toward the characterization of upper tropospheric clouds using Atmospheric Infrared Sounder and Microwave Limb Sounder observations

We estimate the accuracy of cloud top altitude (Z) retrievals from the Atmospheric Infrared Sounder (AIRS) and Advanced Microwave Sounding Unit (AMSU) observing suite (ZA) on board the Earth Observing System Aqua platform. We compare ZA with coincident measurements of Z derived from the micropulse lidar and millimeter wave cloud radar at the Atmospheric Radiation Measurement (ARM) program sites of Nauru and Manus islands (ZARM) and the inferred Z from vertically resolved Microwave Limb Sounder (MLS) ice water content (IWC) retrievals. The mean difference in ZA minus ZARM plus or minus one standard deviation ranges from −2.2 to 1.6 km ± 1.0 to 4.2 km for all cases of AIRS effective cloud fraction (f A) > 0.15 at Manus Island using the cloud radar only. The range of mean values results from using different approaches to determine ZARM, day/night differences, and the magnitude of f A; the variation about the mean decreases for increasing values of f A. Analysis of ZARM from the micropulse lidar at Nauru Island for cases restricted to 0.05 ≤ f A ≤ 0.15 indicates a statistically significant improvement in ZA − ZARM over the cloud radar-derived values at Manus Island. In these cases the ZA − ZARM difference is −1.1 to 2.1 km ± 3.0 to 4.5 km. These results imply that the operational ZA is quantitatively useful for constraining cirrus altitude despite the nominal 45 km horizontal resolution. Mean differences of cloud top pressure (PCLD) inferred from coincident AIRS and MLS ice water content (IWC) retrievals depend upon the method of defining AIRS PCLD (as with the ARM comparisons) over the MLS spatial scale, the peak altitude and maximum value of MLS IWC, and f A. AIRS and MLS yield similar vertical frequency distributions when comparisons are limited to f A > 0.1 and IWC > 1.0 mg m−3. Therefore the agreement depends upon the opacity of the cloud, with decreased agreement for optically tenuous clouds. Further, the mean difference and standard deviation of AIRS and MLS PCLD are highly dependent on the MLS tangent altitude. For MLS tangent altitudes greater than 146 hPa, the strength of the limb technique, the disagreement becomes statistically significant. This implies that AIRS and MLS “agree” in a statistical sense at lower tangent altitudes and “disagree” at higher tangent altitudes. These results provide important insights on upper tropospheric cloudiness as observed by nadir-viewing AIRS and limb-viewing MLS.

Wu, Dong L.↗

1H and 13C NMR Chemical Shift Assignments and Conformational Analysis for the Two Diastereomers of the Vitamin K Epoxide Reductase Inhibitor Brodifacoum

Proton and 13C NMR chemical shift assignments and 1H-1H scalar couplings for the two diastereomers of the vitamin K epoxide reductase (VKOR) inhibitor brodifacoum have been determined from acetone solutions containing both diastereomers. Data were obtained from homo- and heteronuclear correlation spectra acquired at 1H frequencies of 750 and 900 MHz over a 268-303 K temperature range. Conformations inferred from scalar coupling and 1-D NOE measurements exhibit large differences between the diastereomers. Pacific Northwest National Laboratory is operated by Battelle for the US Department of Energy.

59 BASIC BIOLOGICAL SCIENCES↗

Deep Learning for Subsurface Flow: A Comparative Study of U‐Net, Fourier Neural Operators, and Transformers in Underground Hydrogen Storage

Subsurface flow research is essential for the sustainable management of natural resources and the environment. Deep learning (DL) has significantly advanced this field by developing efficient and accurate surrogate models to replace computationally expensive physics‐based simulations. These surrogate models are commonly used to predict the spatiotemporal evolution of state variables, such as gas saturation and reservoir pressure, in heterogeneous geological formations. Despite the various DL models applied to this task, there is a lack of studies systematically comparing their performance. This absence of comparative analysis leads to somewhat arbitrary DL model selection in subsurface flow research, resulting in suboptimal performance and potentially inaccurate predictions. To bridge this gap, we conduct a systematic comparison study of three popular DL architectures—U‐Net, Fourier Neural Operators (FNO), and Segmentation Transformer (SETR)—in surrogate modeling of underground hydrogen storage (UHS). We focus on UHS due to its promise of enhancing clean energy resilience and its cyclic operational conditions that represent common scenarios in various subsurface applications. We evaluate the models based on accuracy, training cost, and inference speed. The comparison shows that U‐Net achieves the highest accuracy, followed by SETR and FNO. Despite its lower accuracy, FNO has the highest inference speed. SETR offers competitive accuracy with the least training memory usage, demonstrating the potential of transformers in learning subsurface flow. Our results provide guidance for selecting DL models for surrogate modeling in a wide range of subsurface flow problems.

42 ENGINEERING↗

Parameter estimation for binary neutron-star coalescences with realistic noise during the advanced LIGO era

Advanced ground-based gravitational-wave (GW) detectors begin operation imminently. Their intended goal is not only to make the first direct detection of GWs, but also to make inferences about the source systems. Binary neutron-star mergers are among the most promising sources. We investigate the performance of the parameter-estimation (PE) pipeline that will be used during the first observing run of the Advanced Laser Interferometer Gravitational-wave Observatory (aLIGO) in 2015: we concentrate on the ability to reconstruct the source location on the sky, but also consider the ability to measure masses and the distance. Accurate, rapid sky localization is necessary to alert electromagnetic (EM) observatories so that they can perform follow-up searches for counterpart transient events. We consider PE accuracy in the presence of non-stationary, non-Gaussian noise. We find that the character of the noise makes negligible difference to the PE performance at a given signal-to-noise ratio. The source luminosity distance can only be poorly constrained, since the median 90% (50%) credible interval scaled with respect to the true distance is 0.85 (0.38). However, the chirp mass is well measured. Our chirp-mass estimates are subject to systematic error because we used gravitational-waveform templates without component spin to carry out inference on signals with moderate spins, but the total error is typically less than 10{sup −3} M{sub ⊙}. The median 90% (50%) credible region for sky localization is ∼600 deg{sup 2} (∼150 deg{sup 2}), with 3% (30%) of detected events localized within 100 deg{sup 2}. Early aLIGO, with only two detectors, will have a sky-localization accuracy for binary neutron stars of hundreds of square degrees; this makes EM follow-up challenging, but not impossible.

79 ASTRONOMY AND ASTROPHYSICS↗

From direct detection to relic abundance: the case of proton-philic spin-dependent inelastic Dark Matter

We discuss strategies to make inferences on the thermal relic abundance of a Weakly Interacting Massive Particle (WIMP) when the same effective dimension-six operator that explains an experimental excess in direct detection is assumed to drive decoupling at freeze-out, and apply them to the explicit scenario of WIMP inelastic up-scattering with spin-dependent couplings to protons (proton-philic Spin-dependent Inelastic Dark Matter, pSIDM), a phenomenological set-up containing two Dark Matter (DM) particles χ{sub 1} and χ{sub 2} with masses m {sub χ}= m {sub χ{sub 1}} and m {sub χ{sub 2}}= m {sub χ}+δ that we have shown in a previous paper to explain the DAMA effect in compliance with the constraints from other detectors. We also update experimental constraints on pSIDM, extend the analysis to the most general spin-dependent momentum-dependent interactions allowed by non-relativistic Effective Field Theory (EFT), and consider for the WIMP velocity distribution in our Galaxy f ( v ) both a halo-independent approach and a standard Maxwellian. Under these conditions we find that the DAMA effect can be explained in terms of the particle χ{sub 1} in compliance with all the other constraints for all the analyzed EFT couplings and also for a Maxwellian f ( v ). As far as the relic abundance is concerned, we show that the problem of calculating it by using direct detection data to fix the model parameters is affected by a strong sensitivity on f ( v ) and by the degeneracy between the WIMP local density ρ{sub χ} and the WIMP-nucleon scattering cross section, since ρ{sub χ} must be rescaled with respect to the observed DM density in the neighborhood of the Sun when the calculated relic density Ω is smaller than the observed one Ω{sub 0}. As a consequence, a DM direct detection experiment is not directly sensitive to the physical cut-off scale of the EFT, but on some dimensional combination that does not depend on the actual value of Ω. However, such degeneracy can be used to develop a consistency test on the possibility that the WIMP is a thermal relic in the first place. When we apply it to the pSIDM scenario we find that only a WIMP with the standard spin-dependent interaction O=χ-bar {sub 1γ}{sup μ}γ{sup 5χ{sub 2}} q-bar γ{sub μ}γ{sup 5} q + h.c. with quarks can be a thermal relic for, approximately, 10 GeV\;≲ m {sub χ}≲ 16 GeV, 17 keV\;≲ δ≲ 28 keV, and a large uncertainty on Ω, 6× 10{sup −7}Ω{sub 0≲} Ω ≲ Ω{sub 0}. In order for the scenario to work the WIMP galactic velocity distribution must depart from a Maxwellian. Moreover, all the χ{sub 2} states must have already decayed today, and this requires some additional mechanism besides that provided by the O operator.

79 ASTRONOMY AND ASTROPHYSICS↗

Structure-aware Initialization via Numerical Continuation and Informed Priors

Scientific machine learning (SciML) often operates in ill-conditioned, weakly identifiable regimes due to limited data or indirect observations. In such settings, optimization and inference are highly sensitive to the starting point, making initialization--often under-reported--a consequential degree of freedom. Random initialization is not a neutral default as it induces an implicit prior over candidate solutions and can systematically bias the result, producing large run-to-run variability. Here, we formalize this view by treating initialization as a hidden confounder in SciML and develop a unifying theory for structure-aware initialization via numerical continuation, constructing warm starts from related problem instances. Across representative tasks, including physics-informed neural networks, maximum likelihood estimation, and variational inference, warm starts have been shown to consistently reduce optimization effort and improve reliability.

Data integrity↗

epi_inference v. 1.0

SAND2021-0503 O The epi_inference software supports the prediction of disease transmission parameters from data. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.

Hart, William↗

An expert system for scheduling requests for communications links between TDRSS and ERBS

An ERBS-TDRSS Contact Planning System (ERBS-TDRSS CPS) is described which uses a graphics interface and the NASA Transportable Inference Engine. The procedure involves transfer of the ERBS-TDRSS Ground Track Orbit Prediction data to the ERBS flight operations area, where the ERBS-TDRSS CPS automatically generates requests for TDRSS service. As requested events are rejected, alternative context sensitive strategies are employed to generate new requested events until a schedule is completed. A report generator builds schedule requests for separate ERBS-TDRSS contacts.

Mclean, David R.↗

Cloud Power Spectra-Dependence on Solar Zenith Angle and Wavelength, Implications for Cloud Optical Property Retrievals

Scale breaks (spatial scales at which power-law exponent changes occur) observed in Landsat radiances have proven to be useful indicators of radiative interactions, and have aided the development of improved techniques in the remote sensing of clouds. This work extends previous theoretical studies to absorbing wavelengths by using both Landsat Thematic Mapper (TM) observations and Monte Carlo (MC) simulations to infer the systematic dependencies of power spectral shape on cloud characteristics, illumination conditions, and wavelength. We show that MC simulations operating on a simple fractal model of horizontally inhomogeneous clouds produce power spectra that qualitatively resemble observed spectra. We also show that the decrease in the spectra power-law exponent seen at intermediate scales (referred to as "roughening") as the Sun becomes more oblique is more pronounced at absorbing wavelengths. An automated procedure designed to detect the small scale break location is unable to find systematic differences between TM Band 4 and Band 7, despite the fact that MC simulations point to systematic differences in horizontal fluxes. The effect of these qualitative characteristics of the spatial spectra on the retrieval of cloud optical properties is examined by comparing power spectra of nadir radiances with power spectra of optical properties retrieved using either traditional Independent Pixel Approximation approaches or modifications based on normalized radiance indices and the inverse Non-local Independent Pixel Approximation. Assuming that the actual cloud properties follow perfect scaling behavior at all scales, we show the improvement of the proposed retrieval modifications.

Oreopoulos, Lazaros↗

Explanation Capabilities for Behavior-Based Robot Control

A recent study that evaluated issues associated with remote interaction with an autonomous vehicle within the framework of grounding found that missing contextual information led to uncertainty in the interpretation of collected data, and so introduced errors into the command logic of the vehicle. As the vehicles became more autonomous through the activation of additional capabilities, more errors were made. This is an inefficient use of the platform, since the behavior of remotely located autonomous vehicles didn't coincide with the "mental models" of human operators. One of the conclusions of the study was that there should be a way for the autonomous vehicles to describe what action they choose and why. Robotic agents with enough self-awareness to dynamically adjust the information conveyed back to the Operations Center based on a detail level component analysis of requests could provide this description capability. One way to accomplish this is to map the behavior base of the robot into a formal mathematical framework called a cost-calculus. A cost-calculus uses composition operators to build up sequences of behaviors that can then be compared to what is observed using well-known inference mechanisms.

Huntsberger, Terrance L.↗

The GLAS Algorithm Theoretical Basis Document for Laser Footprint Location (Geolocation) and Surface Profiles

This ATBD summarizes (and links with other ATBDs) the elements used to obtain the geolocated GLAS laser spot location, with respect to the Earth Center of Mass. Because of the approach used, the reference frame used to express the geolocation is linked to the reference frame used for POD and PAD, which are related to the ITRF. The geolocated spot coordinates (which includes the elevation or height, with respect to an adopted reference ellipsoid) is the inferred position of the laser spot, since the spot location is not directly measured. This document also summarizes the GLAS operation time periods.

Shutz, Bob E.↗