Search NASA⌕ Search

SEARCH · Search NASA

Results for “Convolution”

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 1,711 records · Page 95

Leveraging Natural Language Processing and Generative Models in Molecular Chemistry: Property Prediction and Novel Compound Generation

The accurate prediction of molecular properties is important for the rational design and the advancement of green chemistry and sustainable materials research. However, the predictive power of traditional computational chemistry methods is limited due to computational restrictions. Here, in this study, we examine an alternative approach to the accurate prediction of properties of organic compounds: natural language processing (NLP)-based molecular embedding. Using viscosity, partition coefficient (log P), and enthalpy of vaporization as test properties through a survey of comprehensive datasets comprising 5695 data points for viscosity, 25 870 data points for log P, and 2296 data points for enthalpy of vaporization. These are important properties for the design of greener, safer, and sustainable chemical processes. Models were trained using NLP methods such as Mol2vec and fine-tuned ChemBERTa, and results were compared with traditional input featurization techniques such as Morgan fingerprints and quantum chemistry derived sigma profiles and DFT features. Among the various machine learning models, Mol2vec demonstrated superior predictive capabilities, achieving the highest correlation coefficient (R 2 = 0.945) and lowest RMSE (0.106 mPa s) for viscosity, as well as high accuracy for log P and enthalpy of vaporization predictions. These findings establish the Mol2vec featurization technique, graph-convolutional neural networks (GCNN), and fine-tuned ChemBERTa model as powerful tools for predictive modeling of organic compounds properties, offering a significant improvement over previously used featurization techniques and opening up strategies for very-high-throughput computational screening. Finally, we integrated ML models with hybrid language-model-based generative adversarial networks (LM-GAN) to generate novel molecular sequences with desirable properties for different research applications. The ability to computationally design solvents with lower viscosity, lower log P, and lower enthalpy of vaporization offers a data-driven route to accelerating the discovery of sustainable alternatives to traditionally toxic solvents.

ChemBERTa↗

An Adaptive Fast Multipole Boundary Element Method for Poisson-Boltzmann Electrostatics

The numerical solution of the Poisson Boltzmann (PB) equation is a useful but a computationally demanding tool for studying electrostatic solvation effects in chemical and biomolecular systems. Recently, we have described a boundary integral equation-based PB solver accelerated by a new version of the fast multipole method (FMM). The overall algorithm shows an order N complexity in both the computational cost and memory usage. Here, we present an updated version of the solver by using an adaptive FMM for accelerating the convolution type matrix-vector multiplications. The adaptive algorithm, when compared to our previous nonadaptive one, not only significantly improves the performance of the overall memory usage but also remarkably speeds the calculation because of an improved load balancing between the local- and far-field calculations. We have also implemented a node-patch discretization scheme that leads to a reduction of unknowns by a factor of 2 relative to the constant element method without sacrificing accuracy. As a result of these improvements, the new solver makes the PB calculation truly feasible for large-scale biomolecular systems such as a 30S ribosome molecule even on a typical 2008 desktop computer.

97 MATHEMATICS AND COMPUTING↗

Superresolved Motions of Single Molecular Catalysts during Polymerization Show Wide Distributions

The motion of single molecular ruthenium catalysts during and after single turnover events of ring-opening metathesis polymerization is imaged through single-molecule superresolution tracking with a positional accuracy of ±32 nm. This tracking is achieved through the real-time incorporation of spectrally tagged monomer units into active polymer chain ends during living polymerization; thus, by design, only active-catalyst motion is detected and imaged, without convolution by inactive catalysts. The catalysts show diverse individualistic diffusive behaviors with respect to time that persist for up to 20 s. Catalysts occupy three mobility populations: quasi-stationary (23%), intermediate (53%, 65 nm), and large (24%, 145 nm) step sizes. Differences in catalyst mobility populations also exist between individual aggregates (p < 0.001). Such differential motion indicates widely different local catalyst microenvironments during the catalytic turnover. These mobility differences are uniquely observable through single-catalyst microscopy and are not measurable through traditional ensemble analytical techniques for characterizing the behavior of molecular catalysts, such as nuclear magnetic resonance spectroscopy. The measured distributions of active molecular catalyst motions would not be readily predictable through modeling or first-principles, and the range likely impacts individual catalyst turnover rate and selectivity. This range plausibly contributes to property distributions observable in bulk polymers, such as molecular weight polydispersity (e.g., 1.9 in this system), leading to a revised understanding of the mechanistic, microscale origins of macroscale polymer properties.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Water Clustering Modulates Activity and Enables Hydrogenated Product Formation during Carbon Monoxide Electroreduction in Aprotic Media

Water solvation plays a critical role in a wide range of electrochemical transformations, but its role is often convoluted since water is typically used as both a solvent and a proton source. Here, we experimentally control water speciation and activity using aprotic solvent media during the carbon monoxide reduction reaction (CORR). Remarkably, we show that aprotic solvents that support microheterogeneous water–water clusters lead to significant amounts of CORR products (methane and ethylene) with a maximum ethylene Faradaic efficiency of 22% in acetonitrile (χ H 2 O = 0.2). In contrast, microhomogeneous systems–where water integrates into the solvents’ intermolecular binding network and has lower activity–primarily support the undesired hydrogen evolution reaction (HER). As a result, insights gained expand our understanding of water activity and nonaqueous electrolyte design for other important transformation reactions beyond CO reduction, such as CO2RR and HER.

Electrolytes↗

Influence of Adsorption Site and Wavelength on the Photodesorption of NO from the (Fe,Cr) 3 O 4 (111) Mixed Oxide Surface

The photochemical properties of nitric oxide on a mixed oxide single crystal surface was examined in ultrahigh vacuum (UHV) using temperature programmed desorption (TPD), photon stimulated desorption (PSD) and low energy electron diffraction (LEED). The mixed oxide was a 75% Fe and 25% Cr corundum (0001) oxide film prepared on an α-Al2O3(0001) crystal, however its surface became terminated with a magnetite-like (111) structure after sputter/anneal cleaning, leading to a surface designated of (Fe,Cr)3O4(111). TPD of NO from the (Fe,Cr)3O4(111) surface revealed three chemisorbed states at 220, ~315 and 370 K assigned to NO binding at Fe3+, Cr3+ and Fe2+ sites, respectively. No significant thermal chemistry of NO was detected. NO photodesorption, the primary photochemical pathway in UHV, was sensitive to the adsorption site, with rates at the three adsorption sites following the trend: Fe3+ > Fe2+ > Cr3+. Multiexponential rate behavior seen in the overall NO PSD spectra was linked directly to site heterogeneity being manifested as a convolution of the individual NO photodesorption rates at the three types of surface sites. The photodesorption rate with UV light (365 nm) was ~10 times greater than that in the visible, but the per-photon rates across the visible spectrum (from 460 to 630 nm) were independent of the wavelength, which is suggestive of localized photon absorption at the adsorption site. Results in this study demonstrate that the adsorption site plays a critical role in determining photochemical rates on complex oxide surfaces. This work was supported by the US Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences & Biosciences. Pacific Northwest National Laboratory (PNNL) is a multi-program national laboratory operated for DOE by Battelle. The research was performed using EMSL, a national scientific user facility sponsored by the Department of Energy's Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory.

36 MATERIALS SCIENCE↗

Simulation of chemical reaction via particle tracking: Diffusion‐limited versus thermodynamic rate‐limited regimes

Chemical reactions may be simulated without regard to local concentrations by applying simple probabilistic rules of particle interaction and combination. The forward reaction A + B -> C is coded by calculating the probability that any A and B particles will occupy the same volume over some time interval. This becomes a convolution of the location densities of the two particles. The backward reaction is a simple exponential decay of C particles into A and B particles. When the mixing of reactants is not a limiting process, the classical thermodynamic reaction rates are reproduced. When low mixing (as by diffusion) limits the reaction probabilities, the reaction rates drop significantly, including the rate of approach to global equilibrium. At long enough times, the law of mass action is reproduced exactly in the mean, with some irreducible deviation in the local equilibrium saturations (the equilibrium constant divided by the mass action expression) away from unity. The saturation variability is not sensitive to numerical parameters but depends strongly on how far from equilibrium the system is initiated. This is simply due to a relative paucity of particles of some species as the reaction moves far to one side or the other.

58 GEOSCIENCES↗

Probing Slow Earthquakes With Deep Learning

Abstract Slow earthquakes may trigger failure on neighboring locked faults that are stressed sufficiently to break, and slow slip patterns may evolve before a nearby great earthquake. However, even in the clearest cases such as Cascadia, slow earthquakes and associated tremor have only been observed in intermittent and discrete bursts. By training a convolutional neural network to detect known tremor on a single seismic station in Cascadia, we isolate and identify tremor and slip preceding and following known larger slow events. The deep neural network can be used for the detection of quasi‐continuous tremor, providing a proxy that quantifies the slow slip rate. Furthermore, the model trained in Cascadia recognizes tremor in other subduction zones and also along the San Andreas Fault at Parkfield, suggesting a universality of waveform characteristics and source processes, as posited from experiments and theory.

58 GEOSCIENCES↗

Using Lightning Flashes to Image Thunderclouds

We report the optical energy emitted by lightning flashes interacts with the surrounding cloud medium through scattering and absorption. The optical signals recorded by space-based lightning imagers describe a convolution of lightning flash energetics and radiative transfer effects in the intervening cloud layer. A thundercloud imaging technique is presented that characterizes cloud regions based on how they are illuminated by lightning. This technique models the spatial distribution of optical energy in radiant lightning pulses to determine whether and to what extent each illuminated cloud pixel behaves like a homogeneous planar cloud layer. A gridded product is constructed that differentiates flashes that illuminate convective cells from stratiform flashes with long horizontal channels and anvil flashes whose optical emissions reflect off of nearby cloud surfaces. Lastly, producing this imagery with a rolling 15-min window allows us to visualize changes in convection with a rapid (20 s) update cycle.

58 GEOSCIENCES↗

A Benchmark to Test Generalization Capabilities of Deep Learning Methods to Classify Severe Convective Storms in a Changing Climate

Abstract This is a test case study assessing the ability of deep learning methods to generalize to a future climate (end of 21st century) when trained to classify thunderstorms in model output representative of the present‐day climate. A convolutional neural network (CNN) was trained to classify strongly rotating thunderstorms from a current climate created using the Weather Research and Forecasting model at high‐resolution, then evaluated against thunderstorms from a future climate and found to perform with skill and comparatively in both climates. Despite training with labels derived from a threshold value of a severe thunderstorm diagnostic (updraft helicity), which was not used as an input attribute, the CNN learned physical characteristics of organized convection and environments that are not captured by the diagnostic heuristic. Physical features were not prescribed but rather learned from the data, such as the importance of dry air at mid‐levels for intense thunderstorm development when low‐level moisture is present (i.e., convective available potential energy). Explanation techniques also revealed that thunderstorms classified as strongly rotating are associated with learned rotation signatures. Results show that the creation of synthetic data with ground truth is a viable alternative to human‐labeled data and that a CNN is able to generalize a target using learned features that would be difficult to encode due to spatial complexity. Most importantly, results from this study show that deep learning is capable of generalizing to future climate extremes and can exhibit out‐of‐sample robustness with hyperparameter tuning in certain applications.

54 ENVIRONMENTAL SCIENCES↗

Evaluation of CloudSat Radiative Kernels Using ARM and CERES Observations and ERA5 Reanalysis

Despite the widespread use of the radiative kernel technique for studying radiative feedbacks and radiative forcings, there has not been any systematic, observation-based validation of the radiative kernel method. Here, we utilize observed and reanalyzed radiative fluxes and atmospheric profiles from the Atmospheric Radiation Measurement (ARM) program and ERA5 reanalysis to assess a set of observation-based radiative kernels from CloudSat for six ARM sites. The CloudSat radiative kernels, convoluted with the ERA5 state variables, can almost perfectly reconstruct the monthly anomalies of shortwave (SW) and longwave (LW) radiative fluxes in ERA5 at the surface (SFC) and top-of-atmosphere (TOA) with correlations significantly being greater than 0.95. The biases of kernel-estimated flux anomalies calculated using the ARM-observed state variables can be more than twice as large when compared with the ARM-observed surface flux anomalies and Clouds and Earth’s Radiant Energy System (CERES) observed anomalies at the TOA. Generally, clouds contribute to most (>60%) of the variance of flux anomalies at Southern Great Plain (SGP), Tropical Western Pacific (TWP), and Eastern North Atlantic (ENA), and surface albedo dominates (>69%) the variance of SW flux anomalies at North Slope of Alaska (NSA). Furthermore, the radiative kernels exhibit the lowest correlation (r ~ [0.55,0.85]) when reconstructing SFC LW flux anomalies at SGP, TWP, and ENA, whose biases are related to the possibility that the kernels may not fully capture the characteristics associated with MJO and ENSO at TWP and the presence of clouds at SGP and ENA.

54 ENVIRONMENTAL SCIENCES↗

A Moist Physics Parameterization Based on Deep Learning

Abstract Current moist physics parameterization schemes in general circulation models (GCMs) are the main source of biases in simulated precipitation and atmospheric circulation. Recent advances in machine learning make it possible to explore data‐driven approaches to developing parameterization for moist physics processes such as convection and clouds. This study aims to develop a new moist physics parameterization scheme based on deep learning. We use a residual convolutional neural network (ResNet) for this purpose. It is trained with 1‐year simulation from a superparameterized GCM, SPCAM. An independent year of SPCAM simulation is used for evaluation. In the design of the neural network, referred to as ResCu, the moist static energy conservation during moist processes is considered. In addition, the past history of the atmospheric states, convection, and clouds is also considered. The predicted variables from the neural network are GCM grid‐scale heating and drying rates by convection and clouds, and cloud liquid and ice water contents. Precipitation is derived from predicted moisture tendency. In the independent data test, ResCu can accurately reproduce the SPCAM simulation in both time mean and temporal variance. Comparison with other neural networks demonstrates the superior performance of ResNet architecture. ResCu is further tested in a single‐column model for both continental midlatitude warm season convection and tropical monsoonal convection. In both cases, it simulates the timing and intensity of convective events well. In the prognostic test of tropical convection case, the simulated temperature and moisture biases with ResCu are smaller than those using conventional convection and cloud parameterizations.

54 ENVIRONMENTAL SCIENCES↗

Numerical and Analytical Modeling of Flow Partitioning in Partially Saturated Fracture Networks

Infiltration processes in fractured-porous media remain a crucial, yet not very well understood component of recharge and vulnerability assessment. Under partially-saturated conditions flows in fractures, percolating fracture networks, and fault zones contribute to the fastest spectrum of infiltration velocities via preferential pathways. Specifically, the partitioning dynamics at fracture intersections determine the magnitude of flow fragmentation into vertical and horizontal components and hence the bulk flow velocity and dispersion of fracture networks. In this work, we derive an approximate analytical solution for the partitioning process and validate it using smoothed particle hydrodynamics simulations. The developed transfer function is conceptually based on simulation results and laboratory experiments carried out in previous works. It allows to efficiently simulate flow through fracture networks with simple cubic structure and arbitrary number of fractures and aperture sizes via linear response theory and convolution of a given input signal. We derive a non-dimensional bulk flow velocity ($\widetilde{v}$) and dispersion coefficient ($\widetilde{D}$) to characterize fracture networks in terms of dimensionless horizontal and vertical time scales $\tau_m$ and $\tau_0$. The dispersion coefficient is shown to strongly depend on the horizontal time scale and converges towards a constant value of $0.08$ within reasonable fluid and geometrical parameter ranges, while the non-dimensional velocity exhibits a characteristic $\widetilde{v} \sim \tau_m^{-1/2}$ scaling. Given that hydraulic information is often only available at limited places within (fractured-porous) aquifer system, such as boreholes or springs, our study intends to provide a rudimentary analytical concept to potentially reconstruct internal fracture network geometries from external boundary information, e.g., the dispersive properties of discharge (groundwater level fluctuations).

42 ENGINEERING↗

Automated Classification of Plasma Regions Using 3D Particle Energy Distributions

We investigate the properties of the ion sky maps produced by the Dual Ion Spectrometers (DIS) from the Fast Plasma Investigation (FPI). We have trained a convolutional neural network classifier to predict four regions crossed by the Magnetospheric Multiscale Mission (MMS) on the dayside magnetosphere: solar wind, ion foreshock, magnetosheath, and magnetopause using solely DIS spectrograms. The accuracy of the classifier is urn:x-wiley:21699380:media:jgra56742:jgra56742-math-0001%. We use the classifier to detect mixed plasma regions, in particular to find the bow shock regions. A similar approach can be used to identify the magnetopause crossings and reveal regions prone to magnetic reconnection. Data processing through the trained classifier is fast and efficient and thus can be used for classification for the whole MMS database.

79 ASTRONOMY AND ASTROPHYSICS↗

Event‐Based Training in Label‐Limited Regimes

Abstract The distribution of attributes assigned using data on independent sensors for a specific source, for example, magnitude, can be richly descriptive for final event characterization and associated uncertainty. Attribute distributions can also provide powerful context for event characterization in the absence of comprehensive annotation. This work develops a way to leverage distributional information across a set of sensors in the absence of comprehensive annotation as a domain‐informed regularization term applied during gradient‐based learning. The regularization term is the basis of event‐based training which I show can be a powerful semi‐supervised learning (SSL) approach. I first use a simple feed forward neural network and a toy data set to outline how data set structure interacts with the assumptions inherent to many semi‐supervised learning approaches. I then demonstrate the effectiveness of event‐based training using a deep convolutional neural network for seismic event classification in Utah, which increases SSL accuracy from 92% to 97% on event classification with a limited number of training labels.

Linville, Lisa M.↗

Explore Spatio‐Temporal Learning of Large Sample Hydrology Using Graph Neural Networks

Abstract Streamflow forecasting over gauged and ungauged basins play a vital role in water resources planning, especially under the changing climate. Increased availability of large sample hydrology data sets, together with recent advances in deep learning techniques, has presented new opportunities to explore temporal and spatial patterns in hydrological signatures for improving streamflow forecasting. The purpose of this study is to adapt and benchmark several state‐of‐the‐art graph neural network (GNN) architectures, including ChebNet, Graph Convolutional Network (GCN), and GraphWaveNet, for end‐to‐end graph learning. We explicitly represent river basins as nodes in a graph, learn the spatiotemporal nodal dependencies, and then use the learned relations to predict streamflow simultaneously across all nodes in the graph. The efficacy of the developed GNN models is investigated using the Catchment Attributes and MEteorology for Large‐sample Studies (CAMELS) data set under two settings, fixed graph topology (transductive learning), and variable graph topology (inductive learning), with the latter applicable to prediction in ungauged basins (PUB). Results indicate that GNNs are generally robust and computationally efficient, achieving similar or better performance than a baseline model trained using the long short‐term memory (LSTM) network. Further analyses are conducted to interpret the graph learning process at the edge and node levels and to investigate the effect of different model configurations. We conclude that graph learning constitutes a viable machine learning‐based method for aggregating spatiotemporal information from a multitude of sources for streamflow forecasting

Sun, Alexander Y.↗

Potential Vorticity Dynamics Explain How Extratropical Oceans and the Arctic Modulate Wintertime Land–Temperature Variations

Temperature variations across the continental northern hemisphere at the interdecadal scale are thought to be remotely modulated by oceanic internal climate variability and the Arctic. Nevertheless, further elucidating the dynamics is essential for clarifying ongoing debates. We show that potential vorticity (PV) dynamics provide a concise explanation for these teleconnections. Our findings demonstrate that extratropical oceans and the Arctic can remotely modulate the wintertime continental temperature variations by stimulating PV anomalies, which are constrained by climatological PV gradients and jet streams. A causal explanation includes anomalous temperature and precipitation over oceans and the Arctic inducing local PV anomalies via diabatic heating. Subsequently, meridional and downstream advection distributes the anomalous PV to remote land regions where the climatological PV gradients are strong, that is, involving PV fronts, as well as jet streams; therefore, PV fronts and jet streams jointly indicate land regions that are largely and frequently impacted. However, land can also modulate other regions through the same mechanism when variations over land occur in advance. Clear causality depends on which factor is independent, while interactions among those regions may convolute the causality, thereby causing further debates.

54 ENVIRONMENTAL SCIENCES↗

Detecting Large Explosions With Machine Learning Models Trained on Synthetic Infrasound Data

Explosions produce low-frequency acoustic (infrasound) waves capable of propagating globally, but the spatio-temporal variability of the atmosphere makes detecting events difficult. Machine learning (ML) is well-suited to identify the subtle and nonlinear patterns in explosion infrasound signals, but a previous lack of ground-truth data inhibited training of generalized models. We introduce a physics-based method that propagates infrasound sources through realistic atmospheres to create 28,000 synthetic events, which are used to train ML classifiers. A simple artificial neural network and modern temporal convolutional network discriminate synthetic events from background noise with >90% accuracy and, more importantly, successfully identify the majority of real-world explosion signals recorded during the Humming Road Runner experiment. ML models trained entirely on physics-based synthetics advance explosion detection capabilities and make ML more viable to related fields lacking training data.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Predicting Future Laboratory Fault Friction Through Deep Learning Transformer Models

Machine learning models using seismic emissions as input can predict instantaneous fault characteristics such as displacement and friction in laboratory experiments, and slow slip in Earth. Here, we address whether the seismic/acoustic emission (AE) from laboratory experiments contains information about future frictional behavior. The approach uses a convolutional encoder-decoder containing a transformer model in the latent space, similar to models used for natural language processing. We test the model limits using progressively larger AE input time windows and progressively larger output friction time windows. The results demonstrate that very near-term friction predictions are indeed contained in the AE signal, and predictions are progressively worse farther into the future. The future predictions by the model of impending failure in the near-term are remarkably robust. This first effort predicting future fault frictional behavior with machine learning will aid in guiding efforts for applications in Earth.

58 GEOSCIENCES↗