Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep Gaussian processes”

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 109 records · Page 6

Preliminary Preclosure Accident Consequence Analysis for the APM Conceptual Design - 18090

The Nuclear Waste Management Organization (NWMO) is implementing Adaptive Phased Management (APM), Canada's plan for the long-term management of used nuclear fuel. The APM approach encompasses centralized containment and isolation of the used fuel in a Deep Geological Repository (DGR) in a suitable rock formation, such as crystalline rock or sedimentary rock, in an informed and willing host. This paper describes a preliminary analysis of the potential dose consequences to the public under accident conditions associated with the operation of a DGR and related surface processing facilities. These results are for a generic site and considers exposure to a person standing at a potential fenceline location under conservative atmospheric conditions. Radiological dose consequences to the public were calculated for accidents identified in the preliminary hazard identification. The presence or absence of ventilation system HEPA filters was also considered in combination with specific accident scenarios. A Gaussian dispersion model considering the most stable atmospheric condition (Pasquill stability category F) was used to estimate the exposure of a member of the public in the direct plume path at the location of maximum airborne radionuclide concentrations. The public dose was calculated via the inhalation, air immersion and ground exposure pathways. For all accidents considered, the calculated public dose was low, up to about 0.8 mSv for a fall of an UFTP with container breach. The total dose to the public was also calculated at various distances from the Used Fuel Packaging Plant, Ventilation Shaft, and Main Shaft. For all accidents considered, the calculated dose to the public was low, below 1 mSv even at the minimum 100 m distance evaluated. Sensitivity cases were also carried out to determine the effect of stack release height, the effluent exit velocity, and the release orientation on the calculated minimum site boundary distance. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Improving Dark Energy Constraints Using Low-Redshift Large-Scale Structures

The primary goal of this project was to improve constraints on dark energy measurements by improving our ability to extract cosmological information from low redshift large-scale structures. PI Clowe's project's primary aim was to reduce the bias in measurements of the masses of clusters of galaxies to a level where the evolution of the cluster mass function can be used in the Vera Rubin Observatory's Legacy Survey of Space and Time Dark Energy Science Collaboration survey to improve the accuracy of the measurement of dark energy and other cosmological parameters. Co-PI Seo's project studied observational systematics affecting large-scale clustering of galaxies, which will be used to improve dark energy constraints from the Dark Energy Spectroscopic Instrument (DESI). The cluster lensing project employed a series of simulations and observations of clusters of galaxies to test numerous potential systematic errors in cluster mass measurements using weak gravitational lensing as the accuracy of current weak lensing measurements are more than order of magnitude worse than what is required to use clusters of galaxies for accurate determination of dark energy parameters. PI Clowe and group developed and analyzed simulations to test for, and correct biases introduced in, the weak lensing measurement process. Finally, PI Clowe and group developed a method of detecting clusters using galaxy overdensities and applied the method to the BLISS and DES surveys. The success of spectroscopic dark energy mission such as the extended Baryon Oscillation Spectroscopic Survey (eBOSS) and the Dark Energy Spectroscopic Instrument (DESI) will depend on a thorough understanding of various observational systematics in the target density fluctuations that would give rise to spurious, non-cosmological signals. PI Seo and group developed a deep learning, artificial neural network (ANN) technique that modeled and mitigated such effects, aimed at deriving more robust galaxy clustering signals not only for the baryon acoustic oscillation feature and redshift-space distortions but also for primordial non-Gaussianity constraint.

79 ASTRONOMY AND ASTROPHYSICS↗

Impact of Three-Phase Relative Permeability and Hysteresis Models on Forecasts of Storage Associated With CO 2 -EOR

Geological CO 2 sequestration in conjunction with enhanced oil recovery (CO 2 -EOR) includes complex multiphase flow processes compared to CO 2 storage in deep saline aquifers. Two of the most critical factors affecting multiphase flow in CO 2 -EOR are three-phase relative permeability and associated hysteresis, both of which are difficult to measure and are usually represented by numerical interpolation models. The purpose of this research is to improve understanding of (1) the relative impacts of different three-phase relative permeability models and hysteresis models on CO 2 trapping mechanisms, and (2) uncertainty associated with these two factors. Four different three-phase relative permeability models and three hysteresis models were applied to simulations of an active CO 2 -EOR site, the SACROC unit located in western Texas. To eliminate potential bias of deterministic parameters, we utilized a sequential Gaussian simulation technique to generate 50 realizations to describe heterogeneity of porosity and permeability, based on data obtained from well logs and seismic survey. Simulation results of forecasted CO 2 storage suggested that (1) the choice of three-phase relative permeability model and hysteresis model led to noticeable impacts on forecasted CO 2 sequestration capacity; (2) impacts of three-phase relative permeability models and hysteresis models on CO 2 trapping are small during the CO 2 -EOR injection period, and increase during the post-EOR CO 2 injection period; (3) the specific choice of hysteresis model is more important relative to the choice of three-phase relative permeability model; and (4) using the recommended three-phase WAG (Water-Alternating-Gas) hysteresis model may increase the impact of three-phase relative permeability models and uncertainty due to heterogeneity.

42 ENGINEERING↗

Discriminative Dimensionality Reduction using Deep Neural Networks for Clustering of LIGO Data

In this paper, leveraging the capabilities of neural networks for modeling the non-linearities that exist in the data, we propose several models that can project data into a low dimensional, discriminative, and smooth manifold. The proposed models can transfer knowledge from the domain of known classes to a new domain where the classes are unknown. A clustering algorithm is further applied in the new domain to find potentially new classes from the pool of unlabeled data. The research problem and data for this paper originated from the Gravity Spy project which is a side project of Advanced Laser Interferometer Gravitational-wave Observatory (LIGO). The LIGO project aims at detecting cosmic gravitational waves using huge detectors. However non-cosmic, non-Gaussian disturbances known as "glitches", show up in gravitational-wave data of LIGO. This is undesirable as it creates problems for the gravitational wave detection process. Gravity Spy aids in glitch identification with the purpose of understanding their origin. Since new types of glitches appear over time, one of the objective of Gravity Spy is to create new glitch classes. Towards this task, we offer a methodology in this paper to accomplish this.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Accelerating Hydraulic Fracture Imaging by Deep Transfer Learning

Deep transfer learning has a great success story in computer vision (CV), natural language processing, and many other fields. In this communication, we are going to push forward the deep transfer learning to the hydraulic fracture imaging problem by proposing a two-step approach: 1) train a convolutional neural network (CNN) to reconstruct target geometries by a relatively large amount of approximated field patterns generated from a simplified model and 2) fine-tune the top layers of transferred CNN by a small amount of true field patterns generated through a full model. The advantages include the rapid generation of large amount of data through the simplified model and the high reconstruction accuracy through a careful design of the deep transfer learning. Here, the CNN trained through the deep transfer learning can accurately reconstruct the lateral extent and direction of fractures with unseen conductivity and white Gaussian noise, showing a notable acceleration/accuracy enhancement over the previous CNN trained by mixing a small number of true data into the approximated data as data augmentation.

02 PETROLEUM↗

Real-time Intrusion Detection for High-bandwidth Research Networks using Unsupervised Deep Learning

Nowadays, data generated by large-scale scientific experiments are on the scale of petabytes per month. These data are transferred through dedicated high-bandwidth networks across distributed collaboration sites for processing, storage, and analysis. Like general purpose networks, research networks experience attacks. However, monitoring traffic in such high-bandwidth network environments is challenging with conventional deep packet inspection (DPI)- based network intrusion detection systems (NIDS). Recently, machine learning (ML) algorithms have shown potential in detecting anomalous patterns with better generalizability and faster speed. Compared to the general purpose networks, traffic movement in research network is less volatile and more likely to benefit from the ML-based anomaly analysis. We present ThunderSecure, an efficient AI-powered system for detecting anomalies in 100G research networks. ThunderSecure implements a high-throughput packet processing infrastructure using multi-cores and GPUs. It takes time-series statistics from a variety of network measurements and uses 2D Convolutional Neural Network (CNN) to explore both spatial and temporal dependencies in network data stream. Patterns of science data traffic are learned with a one-class neural network, which blends an Adversarial Autoencoder (AAE) with Gaussian mixture density estimation. Testing traffic flows exhibiting significant deviations from the learned baseline of normality are marked as anomalies. We trained ThunderSecure on hundreds of billions of science data packets recorded from a 100G research network at Fermi National Accelerator Laboratory. The detection performance was evaluated on traffic captured from the same research network days and weeks after the training with different types of attack flows injected. To the best of our knowledge, this is the first ML work in the area of network anomaly detection that has been validated on such extreme scale datasets. Results show that ThunderSecure can recognize science data traffic that are captured long after the training and make nearly certain detection on those with anomalous flows injected, even in the case when the anomaly-to-normal mixing ratio is 0.1%.

Gong, Qian↗

Physics-Informed Generative Adversarial Networks for Stochastic Differential Equations

We developed a new class of physics-informed generative adversarial networks (PI-GANs) to solve forward, inverse, and mixed stochastic problems in a unified manner based on a limited number of scattered measurements. Unlike standard GANs relying solely on data for training, here we encoded into the architecture of GANs the governing physical laws in the form of stochastic differential equations (SDEs) using automatic differentiation. In particular, we applied Wasserstein GANs with gradient penalty (WGAN-GP) for its enhanced stability compared to vanilla GANs. We first tested WGAN-GP in approximating Gaussian processes of different correlation lengths based on data realizations collected from simultaneous reads at sparsely placed sensors. We obtained good approximation of the generated stochastic processes to the target ones even if there is a mismatch between the input noise dimensionality and the effective dimensionality of the target stochastic processes. We also studied the overfitting issue for both the discriminator and the generator, and we found that overfitting occurs also in the generator in addition to the discriminator as previously reported. Subsequently, we considered the solution of elliptic SDEs requiring approximations of three stochastic processes, namely the solution, the forcing, and the diffusion coefficient. Here again, we assumed data collected from simultaneous reads at a limited number of sensors for the multiple stochastic processes. Furthermore, three generators were used for the PI-GANs: two of them were feed forward deep neural networks (DNNs), while the other one was the neural network induced by the SDE. For the case where we have one group of data, we employed one feed forward DNN as the discriminator, while for the case of multiple groups of data we employed multiple discriminators in PI-GANs. We solved forward, inverse, and mixed problems without changing the framework of PI-GANs, obtaining both the means and the standard deviations of the stochastic solution and the diffusion coefficient in good agreement with benchmarks. In this work, we have demonstrated the effectiveness of PI-GANs in solving SDEs for about 120 dimensions. In principle, PI-GANs could tackle very high dimensional problems given more sensor data with low-polynomial growth in computational cost.

97 MATHEMATICS AND COMPUTING↗

Surrogate Modeling of Nonlinear Dynamic Systems: A Comparative Study

Surrogate models play a vital role in overcoming the computational challenge in designing and analyzing nonlinear dynamic systems, especially in the presence of uncertainty. This paper presents a comparative study of different surrogate modeling techniques for nonlinear dynamic systems. Four surrogate modeling methods, namely, Gaussian process (GP) regression, a long short-term memory (LSTM) network, a convolutional neural network (CNN) with LSTM (CNN-LSTM), and a CNN with bidirectional LSTM (CNN-BLSTM), are studied and compared. All these model types can predict the future behavior of dynamic systems over long periods based on training data from relatively short periods. The multi-dimensional inputs of surrogate models are organized in a nonlinear autoregressive exogenous model (NARX) scheme to enable recursive prediction over long periods, where current predictions replace inputs from the previous time window. Three numerical examples, including one mathematical example and two nonlinear engineering analysis models, are used to compare the performance of the four surrogate modeling techniques. The results show that the GP-NARX surrogate model tends to have more stable performance than the other three deep learning (DL)-based methods for the three particular examples studied. The tuning effort of GP-NARX is also much lower than its deep learning-based counterparts.

42 ENGINEERING↗

In Pursuit of the FIP Effect in Late-Type Stellar Coronae

Spectral line data for several coronally active stars, in addition to EUVE Deep Survey light curves, have been analysed under this program. Much difficulty has been encountered in the study that has resulted in fewer stars being analysed than had been hoped. The difficulties stemmed from the analysis of low X-ray spectra taken with the ASCA satellite that produced results that are strongly discrepant with respect to the EUVE results. There is no obvious explanation for this, though it appears that analysis of ASCA data systematically underestimate metal abundance in hot plasmas. Consequently, the final emphasis in our analyses has been on EUVE data. Observed line profiles have being fitted in order to measure their fluxes using IDL software specially developed under this and parallel efforts.The observed line profiles deviate from pure gaussian forms, but we have found the benefits of using additional functional forms in the fitting process to be of only very small value for the lines with highest S/N. The resulting line fluxes have being processed in terms of the coronal EM using new techniques. Resulting EM distribution models are being used to finalize metallicity and abundance estimates for the stars in the program. Special account of the influence of missing lines in the spectral models has been taken.

Drake, Jeremy↗

Deep Search for Decaying Dark Matter with XMM-Newton Blank-Sky Observations

Sterile neutrinos with masses in the keV range are well-motivated extensions to the Standard Model that could explain the observed neutrino masses while also making up the dark matter (DM) of the universe. If sterile neutrinos are DM then they may slowly decay into active neutrinos and photons, giving rise to the possibility of their detection through narrow spectral features in astrophysical x-ray data sets. Here, we perform the most sensitive search to date for this and other decaying DM scenarios across the mass range from 5 to 16 keV using archival XMM-Newton data. We reduce 547 Ms of data from both the MOS and PN instruments using observations taken across the full sky and then use this data to search for evidence of DM decay in the ambient halo of the Milky Way. We determine the instrumental and astrophysical baselines with data taken far away from the Galactic Center, and use Gaussian process modeling to capture additional continuum background contributions. No evidence is found for unassociated x-ray lines, leading us to produce the strongest constraints to date on decaying DM in this mass range.

79 ASTRONOMY AND ASTROPHYSICS↗

Advection of a passive vector field by the Gaussian velocity field with finite correlations in time

Using the field theoretic renormalization group technique the model of a passive vector field advected by an incompressible turbulent flow is investigated up to the second order of the perturbation theory (two-loop approximation). The turbulent environment is given by statistical fluctuations of the velocity field that has a Gaussian distribution with zero mean and defined noise with finite correlations in time. Two-loop analysis of all possible scaling regimes in general d-dimensional space is done in the plane of exponents ɛ − η, where ɛ characterizes the energy spectrum of the velocity field in the inertial range E ∝ k{sup 1−2ε}, and η is related to the correlation time at the wave number k which is scaled as k{sup −2+η}. It is shown that the scaling regimes of the present model of vector advection have essentially different properties than the scaling regimes of the corresponding model of passively advected scalar quantity. The results demonstrate the fact that within the present model of passively advected vector field the internal tensor structure of the advected field can have nontrivial impact on the diffusion processes deep inside in the inertial interval of given turbulent flow.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Compression induced phase transition of nematic brush: A mean-field theory study

Responsive behavior of polymer brush to the external compression is one of the most important characters for its application. For the flexible polymer brush, in the case of low grafting density, which is widely studied by the Gaussian chain model based theory, the compression leads to a uniform deformation of the chain. However, in the case of high grafting density, the brush becomes anisotropic and the nematic phase will be formed. The normal compression tends to destroy the nematic order, which leads to a complex responsive behaviors. Under weak compression, chains in the nematic brush are buckled, and the bending energy and Onsager interaction give rise to the elasticity. Under deep compression, the responsive behaviors of the nematic polymer brush depend on the chain rigidity. For the compressed rigid polymer brush, the chains incline to re-orientate randomly to maximize the orientational entropy and its nematic order is destroyed. For the compressed flexible polymer brush, the chains incline to fold back to keep the nematic order. A buckling-folding transition takes place during the compressing process. For the compressed semiflexible brush, the chains are collectively tilted to a certain direction, which leads to the breaking of the rotational symmetry in the lateral plane. These responsive behaviors of nematic brush relate to the properties of highly frustrated worm-like chain, which is hard to be studied by the traditional self-consistent field theory due to the difficulty to solve the modified diffusion equation. To overcome this difficulty, a single chain in mean-field theory incorporating Monte Carlo simulation and mean-field theory for the worm-like chain model is developed in present work. This method shows high performance for entire region of chain rigidity in the confined condition.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Operator learning for predicting multiscale bubble growth dynamics

We report simulating and predicting multiscale problems that couple multiple physics and dynamics across many orders of spatiotemporal scales is a great challenge that has not been investigated systematically by deep neural networks (DNNs). Herein, we develop a framework based on operator regression, the so-called deep operator network (DeepONet), with the long-term objective to simplify multiscale modeling by avoiding the fragile and time-consuming “hand-shaking” interface algorithms for stitching together heterogeneous descriptions of multiscale phenomena. To this end, as a first step, we investigate if a DeepONet can learn the dynamics of different scale regimes, one at the deterministic macroscale and the other at the stochastic microscale regime with inherent thermal fluctuations. Specifically, we test the effectiveness and accuracy of the DeepONet in predicting multirate bubble growth dynamics, which is described by a Rayleigh–Plesset (R–P) equation at the macroscale and modeled as a stochastic nucleation and cavitation process at the microscale by dissipative particle dynamics (DPD). First, we generate data using the R–P equation for multirate bubble growth dynamics caused by randomly time-varying liquid pressures drawn from Gaussian random fields (GRFs). Our results show that properly trained DeepONets can accurately predict the macroscale bubble growth dynamics and can outperform long short-term memory networks. We also demonstrate that the DeepONet can extrapolate accurately outside the input distribution using only very few new measurements. Subsequently, we train the DeepONet with DPD data corresponding to stochastic bubble growth dynamics. Although the DPD data are noisy and we only collect sparse data points on the trajectories, the trained DeepONet model is able to predict accurately the mean bubble dynamics for time-varying GRF pressures. Taken together, our findings demonstrate that DeepONets can be employed to unify the macroscale and microscale models of the multirate bubble growth problem, hence providing new insight into the role of operator regression via DNNs in tackling realistic multiscale problems and in simplifying modeling with heterogeneous descriptions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

The physical mechanism on the threshold voltage temperature stability improvement for GaN HEMTs with pre-fluorination argon treatment

In this paper, a normally-off AlGaN/GaN MIS-HEMT with improved threshold voltage (V{sub TH}) thermal stability is reported with investigations on its physical mechanism. The normally-off operation of the device is achieved from novel short argon plasma treatment (APT) prior to the fluorine plasma treatment (FPT) on Al{sub 2}O{sub 3} gate dielectrics. For the MIS-HEMT with FPT only, its V{sub TH} drops from 4.2 V at room temperature to 0.5 V at 200 °C. Alternatively, for the device with APT-then-FPT process, its V{sub TH} can retain at 2.5 V at 200 °C due to the increased amount of deep-level traps that do not emit electrons at 200 °C. This thermally stable V{sub TH} makes this device suitable for high power applications. The depth profile of the F atoms in Al{sub 2}O{sub 3}, measured by the secondary ion mass spectroscopy, reveals a significant increase in the F concentration when APT is conducted prior to FPT. The X-ray photoelectron spectroscopy (XPS) analysis on the plasma-treated Al{sub 2}O{sub 3} surfaces observes higher composition of Al-F bonds if APT was applied before FPT. The enhanced breaking of Al-O bonds due to Ar bombardment assisted in the increased incorporation of F radicals at the surface during the subsequent FPT process. The Schrödinger equation of Al{sub 2}O{sub x}F{sub y} cells, with the same Al-F compositions as obtained from XPS, was solved by Gaussian 09 molecular simulations to extract electron state distribution as a function of energy. The simulation results show creation of the deeper trap states in the Al{sub 2}O{sub 3} bandgap when APT is used before FPT. Finally, the trap distribution extracted from the simulations is verified by the gate-stress experimental characterization to confirm the physical mechanism described.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Electron energy loss spectroscopy database synthesis and automation of core-loss edge recognition by deep-learning neural networks

Abstract The ionization edges encoded in the electron energy loss spectroscopy (EELS) spectra enable advanced material analysis including composition analyses and elemental quantifications. The development of the parallel EELS instrument and fast, sensitive detectors have greatly improved the acquisition speed of EELS spectra. However, the traditional way of core-loss edge recognition is experience based and human labor dependent, which limits the processing speed. So far, the low signal–noise ratio and the low jump ratio of the core-loss edges on the raw EELS spectra have been challenging for the automation of edge recognition. In this work, a convolutional-bidirectional long short-term memory neural network (CNN-BiLSTM) is proposed to automate the detection and elemental identification of core-loss edges from raw spectra. An EELS spectral database is synthesized by using our forward model to assist in the training and validation of the neural network. To make the synthesized spectra resemble the real spectra, we collected a large library of experimentally acquired EELS core edges. In synthesize the training library, the edges are modeled by fitting the multi-Gaussian model to the real edges from experiments, and the noise and instrumental imperfectness are simulated and added. The well-trained CNN-BiLSTM network is tested against both the simulated spectra and real spectra collected from experiments. The high accuracy of the network, 94.9%, proves that, without complicated preprocessing of the raw spectra, the proposed CNN-BiLSTM network achieves the automation of core-loss edge recognition for EELS spectra with high accuracy.

36 MATERIALS SCIENCE↗

Parametric behaviors of CLUBB in simulations of low clouds in the C ommunity A tmosphere M odel ( CAM )

Abstract In this study, we investigate the sensitivity of simulated low clouds to 14 selected tunable parameters of Cloud Layers Unified By Binormals (CLUBB), a higher‐order closure (HOC) scheme, and four parameters of the Zhang‐McFarlane (ZM) deep convection scheme in the Community Atmosphere Model version 5 (CAM5). A Quasi‐Monte Carlo (QMC) sampling approach is adopted to effectively explore the high‐dimensional parameter space and a generalized linear model is applied to study the responses of simulated cloud fields to tunable parameters. Our results show that the variance in simulated low‐cloud properties (cloud fraction and liquid water path) can be explained by the selected tunable parameters in two different ways: macrophysics itself and its interaction with microphysics. First, the parameters related to dynamic and thermodynamic turbulent structure and double Gaussian closure are found to be the most influential parameters for simulating low clouds. The spatial distributions of the parameter contributions show clear cloud‐regime dependence. Second, because of the coupling between cloud macrophysics and cloud microphysics, the coefficient of the dissipation term in the total water variance equation is influential. This parameter affects the variance of in‐cloud cloud water, which further influences microphysical process rates, such as autoconversion, and eventually low‐cloud fraction. This study improves understanding of HOC behavior associated with parameter uncertainties and provides valuable insights for the interaction of macrophysics and microphysics.

54 ENVIRONMENTAL SCIENCES↗

Optimization and Evaluation of Stochastic Unified Convection Using Single-Column Model Simulations at Multiple Observation Sites

We extend the previously developed stochastic unified convection scheme (UNICON) for shallow convection to deep convection by parameterizing the impact of mesoscale organized flow on updraft properties. The extended stochastic UNICON parameterizes thermodynamic properties of updrafts at the near-surface as a multivariate Gaussian distribution, where the variances of the distribution are the summation of variances from non-organized turbulence and mesoscale organized flow. The distribution of updraft radius is parameterized as a power-law distribution with a scale break which is parameterized as a linear function of the strength of mesoscale organized flow. The proposed parameterization is validated using a series of large-eddy simulations of deep convection. The free parameters introduced in the formulation of stochastic UNICON are optimized using 10 cases of single-column model simulations over the ocean. Stochastic UNICON with the optimized parameters significantly reduces the biases of thermodynamic profiles and surface precipitation rates simulated in the original UNICON for tropical convection cases. The simulation of the variation in anomalies of temperature and moisture associated with the Madden-Julian oscillation is also improved. The overall improvements in simulated thermodynamic profiles are found to be due to the increased heating and drying tendencies by convective processes in stochastic UNICON. An additional simulation of an idealized deep convection case shows that stochastic UNICON produces enhanced cloud variabilities with dependency on updraft radius, indicating its ability to represent the coexistence of shallow and deep convection.

54 ENVIRONMENTAL SCIENCES↗