Search NASA⌕ Search

SEARCH · Search NASA

Results for “Generative modeling”

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 469 records · Page 26

A guide on modeling electrostatics of semiconductor detectors in COMSOL Multiphysics ® for DRiFT

Adding a new semiconductor detector into the detector response function toolkit (DRiFT) requires a model of the electric potential and the electric field. To get an accurate electrostatics model, the geometry of the detector should be modeled as closely to the real geometry as feasible, including the semiconductor materials, doping layers and concentrations, and contacts. Since the electric field and potential are used to calculate the induced signal, the results of the detector response functions are largely influenced by the electrostatics models. Initial electrostatics models were completed using Silvaco, however, this document provides guidance on using COMSOL Mulitphysics ® (COMSOL) to model the detectors to give users additional flexibility. Guidance on using COMSOL and its user interface are mostly left out of this document, but best practices for geometry, modeling methods, and data format/exporting are included to help users generate models that are compatible with DRiFT, prevents bugs, issues, and inaccuracies during charge collection calculations. The following summary provides an overview of the document.

42 ENGINEERING↗

Appendix E: Other NEMS-MP results for the base case and scenarios.

The NEMS-MP model generates numerous results for each run of a scenario. (This model is the integrated National Energy Modeling System [NEMS] version used for the Multi-Path Transportation Futures Study [MP].) This appendix examines additional findings beyond the primary results reported in the Multi-Path Transportation Futures Study: Vehicle Characterization and Scenario Analyses (Reference 1). These additional results are provided in order to help further illuminate some of the primary results. Specifically discussed in this appendix are: (1) Energy use results for light vehicles (LVs), including details about the underlying total vehicle miles traveled (VMT), the average vehicle fuel economy, and the volumes of the different fuels used; (2) Resource fuels and their use in the production of ethanol, hydrogen (H{sub 2}), and electricity; (3) Ethanol use in the scenarios (i.e., the ethanol consumption in E85 vs. other blends, the percent of travel by flex fuel vehicles on E85, etc.); (4) Relative availability of E85 and H2 stations; (5) Fuel prices; (6) Vehicle prices; and (7) Consumer savings. These results are discussed as follows: (1) The three scenarios (Mixed, (P)HEV & Ethanol, and H2 Success) when assuming vehicle prices developed through literature review; (2) The three scenarios with vehicle prices that incorporate the achievement of the U.S. Department of Energy (DOE) program vehicle cost goals; (3) The three scenarios with 'literature review' vehicle prices, plus vehicle subsidies; and (4) The three scenarios with 'program goals' vehicle prices, plus vehicle subsidies. The four versions or cases of each scenario are referred to as: Literature Review No Subsidies, Program Goals No Subsidies, Literature Review with Subsidies, and Program Goals with Subsidies. Two additional points must be made here. First, none of the results presented for LVs in this section include Class 2B trucks. Results for this class are included occasionally in Reference 1. They represent a small, though noticeable, segment of the 'LV plus 2B' market (e.g., a little more than 3% of today's energy use in that market). We generally do not include them in this discussion, simply because it requires additional effort to combine the NEMS-MP results for them with the results for the other LVs. (Where there is an exception, we will indicate so.) Second, where reference is made to E85, the ethanol content is actually 74%. The Energy Information Administration (EIA) assumes that, to address cold-starting issues, the percent of ethanol in E85 will vary seasonally. The EIA uses an annual average ethanol content of 74% in its forecasts. That assumption is maintained in the NEMS-MP scenario runs.

08 HYDROGEN↗

Mining Large Heterogeneous Graphs using Cray s Urika

Pattern discovery and predictive modeling from seemingly related Big Data represented as massive, ad-hoc, heterogeneous networks (e.g., extremely large graphs with complex, possibly unknown structure) is an outstanding problem in many application domains. To address this problem, we are designing graph-mining algorithms capable of discovering relationship-patterns from such data and using those discovered patterns as features for classification and predictive modeling. Specifically, we are: (i) exploring statistical properties, mechanics and generative models of behavior patterns in heterogeneous information networks, (ii) developing novel, automated and scalable graph-pattern discovery algorithms and (iii) applying our relationship-analytics (data science + network science) expertise to domains spanning healthcare to homeland security.

fraud detection↗

Fast and accurate simulations of calorimeter showers with normalizing flows

In this study, we introduce caloflow, a fast detector simulation framework based on normalizing flows. For the first time, we demonstrate that normalizing flows can reproduce many-channel calorimeter showers with extremely high fidelity, providing a fresh alternative to computationally expensive geant4 simulations, as well as other state-of-the-art fast simulation frameworks based on generative adversarial networks (GANs) or variational autoencoders (VAEs). In addition to the usual histograms of physical features and images of calorimeter showers, we introduce a new metric for judging the quality of generative modeling: the performance of a classifier trained to differentiate real from generated images. We show that GAN-generated images can be identified by the classifier with nearly 100% accuracy, while images generated from caloflow are better able to fool the classifier. More broadly, normalizing flows offer several advantages compared to other state-of-the-art approaches (GANs and VAEs), including tractable likelihoods, stable and convergent training, and principled model selection. Normalizing flows also provide a bijective mapping between data and the latent space, which could have other applications beyond simulation, for example, to detector unfolding.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Contrasting Time-Frequency Representations for Unknown Waveform Detection

Identifying unseen electromagnetic waveforms is critical for many applications, like interference management, electronic warfare and spectrum management. Traditionally this is done using statistical methods for anomaly detection, which has evolved to deep learning models for identifying the unseen data, formally termed as open set recognition. Some prior methods use a generative model to emulate open set data, which face challenges in generating synthetic samples for open set while simultaneously selecting an optimal discriminator for accurate classification. To alleviate this issue, we propose a discriminative model that effectively combines time and frequency domain features of communication signals for accurate predictions. We further introduce a cosine similarity loss that makes the domain specific features unique to enhance the prediction rate. Additionally, our model avoids generic feature vectors by extracting class-specific features during training, resulting in improved class representation. The experiment results show that this combined feature approach with cosine loss outperforms single-domain models and improves accuracy by 10% over models without cosine loss.

99 - GENERAL AND MISCELLANEOUS↗

Consistent modelling of wind turbine noise propagation from source to receiver

The unsteady nature of wind turbine noise is a major reason for annoyance. The variation of far-field sound pressure levels is not only caused by the continuous change in wind turbine noise source levels but also by the unsteady flow field and the ground characteristics between the turbine and receiver. To take these phenomena into account, a consistent numerical technique that models the sound propagation from the source to receiver is developed. Large eddy simulation with an actuator line technique is employed for the flow modelling and the corresponding flow fields are used to simulate sound generation and propagation. The local blade relative velocity, angle of attack, and turbulence characteristics are input to the sound generation model. Time-dependent blade locations and the velocity between the noise source and receiver are considered within a quasi-3D propagation model. Long-range noise propagation of a 5 MW wind turbine is investigated. Sound pressure level time series evaluated at the source time are studied for varying wind speeds, surface roughness, and ground impedances within a 2000 m radius from the turbine.

17 WIND ENERGY↗

Hierarchical transfer learning: an agile and equitable strategy for machine-learning interatomic models

Machine-learned interatomic models are growing in popularity due to their ability to afford near quantum-accurate predictions for complex phenomena with orders-of-magnitude greater computational efficiency. However, these models struggle when applied to systems of many element types due to the approximately exponential increase in number of parameters that must be determined. To mitigate this challenge, we present a new hierarchical transfer learning approach that allows the fitting problem to be decomposed into smaller independent and reusable parameter blocks that enable development of explicitly chemically extensible ML-IAM. Application of this strategy is demonstrated for C and N mixtures under conditions ranging from nominally ambient to ~10,000 K and 200 GPa for compositions from 0 to 100% N. Ultimately, this strategy makes model generation for chemically complex systems more tractable and efficient, facilitates comprehensive model validation, and makes ML-IAM development for problems of this nature more accessible to users with limited access to extreme computing infrastructure.

Lindsey, Rebecca K. [Univ. of Michigan, Ann Arbor,↗

FAST VARIABILITY AND MILLIMETER/IR FLARES IN GRMHD MODELS OF Sgr A* FROM STRONG-FIELD GRAVITATIONAL LENSING

We explore the variability properties of long, high-cadence general relativistic magnetohydrodynamic (GRMHD) simulations across the electromagnetic spectrum using an efficient, GPU-based radiative transfer algorithm. We focus on both standard and normal evolution (SANE) and magnetically arrested disk (MAD) simulations with parameters that successfully reproduce the time-averaged spectral properties of Sgr A* and the size of its image at 1.3 mm. We find that the SANE models produce short-timescale variability with amplitudes and power spectra that closely resemble those inferred observationally. In contrast, MAD models generate only slow variability at lower flux levels. Neither set of models shows any X-ray flares, which most likely indicates that additional physics, such as particle acceleration mechanisms, need to be incorporated into the GRMHD simulations to account for them. The SANE models show strong, short-lived millimeter/infrared (IR) flares, with short (≲1 hr) time lags between the millimeter and IR wavelengths, that arise from the combination of short-lived magnetic flux tubes and strong-field gravitational lensing near the horizon. Such events provide a natural explanation for the observed IR flares with no X-ray counterparts.

79 ASTRONOMY AND ASTROPHYSICS↗

ATAT: Astronomical Transformer for time series and Tabular data

Context. The advent of next-generation survey instruments, such as theVera C. RubinObservatory and its Legacy Survey of Space and Time (LSST), is opening a window for new research in time-domain astronomy. The Extended LSST Astronomical Time-Series Classification Challenge (ELAsTiCC) was created to test the capacity of brokers to deal with a simulated LSST stream. Aims. Our aim is to develop a next-generation model for the classification of variable astronomical objects. We describe ATAT, the Astronomical Transformer for time series And Tabular data, a classification model conceived by the ALeRCE alert broker to classify light curves from next-generation alert streams. ATAT was tested in production during the first round of the ELAsTiCC campaigns. Methods. ATAT consists of two transformer models that encode light curves and features using novel time modulation and quantile feature tokenizer mechanisms, respectively. ATAT was trained on different combinations of light curves, metadata, and features calculated over the light curves. We compare ATAT against the current ALeRCE classifier, a balanced hierarchical random forest (BHRF) trained on human-engineered features derived from light curves and metadata. Results. When trained on light curves and metadata, ATAT achieves a macro F1 score of 82.9 ± 0.4 in 20 classes, outperforming the BHRF model trained on 429 features, which achieves a macro F1 score of 79.4 ± 0.1. Conclusions. The use of transformer multimodal architectures, combining light curves and tabular data, opens new possibilities for classifying alerts from a new generation of large etendue telescopes, such as theVera C. RubinObservatory, in real-world brokering scenarios.

Astronomy & Astrophysics↗

Systematically improvable multiscale solver for correlated electron systems

The development of numerical methods capable of simulating realistic materials with strongly correlated electrons, with controllable errors, is a central challenge in quantum many-body physics. Here we describe a framework for a general multiscale method based on embedding a self-energy of a strongly correlated subsystem into a self-energy generated by a method able to treat large weakly correlated systems approximately. As an example, we present the embedding of an exact diagonalization self-energy into a self-energy generated from self-consistent second-order perturbation theory. Using a quantum impurity model, generated from a cluster dynamical mean field approximation to the two-dimensional Hubbard model, as a benchmark, we illustrate that our method allows us to obtain accurate results at a fraction of the cost of typical Monte Carlo calculations. We test the method in multiple regimes of interaction strengths and dopings of the model. Here, the general embedding framework we present avoids difficulties such as double counting corrections, frequency-dependent interactions, or vertex functions. As it is solely formulated at the level of the single-particle Green's function, it provides a promising route for the simulation of realistic materials that are currently difficult to study with other methods.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

MO-F-CAMPUS-T-04: Development and Evaluation of a Knowledge-Based Model for Treatment Planning of Lung Cancer Patients Using Stereotactic Body Radiotherapy (SBRT)

Purpose: To describe the development of a knowledge-based treatment planning model for lung cancer patients treated with SBRT, and to evaluate the model performance and applicability to different planning techniques and tumor locations. Methods: 105 lung SBRT plans previously treated at our institution were included in the development of the model using Varian’s RapidPlan DVH estimation algorithm. The model was trained with a combination of IMRT, VMAT, and 3D–CRT techniques. Tumor locations encompassed lesions located centrally vs peripherally (43:62), upper vs lower (62:43), and anterior vs posterior lobes (60:45). The model performance was validated with 25 cases independent of the training set, for both IMRT and VMAT. Model generated plans were created with only one optimization and no planner intervention. The original, general model was also divided into four separate models according to tumor location. The model was also applied using different beam templates to further improve workflow. Dose differences to targets and organs-at-risk were evaluated. Results: IMRT and VMAT RapidPlan generated plans were comparable to clinical plans with respect to target coverage and several OARs. Spinal cord dose was lowered in the model-based plans by 1Gy compared to the clinical plans, p=0.008. Splitting the model according to tumor location resulted in insignificant differences in DVH estimation. The peripheral model decreased esophagus dose to the central lesions by 0.5Gy compared to the original model, p=0.025, and the posterior model increased dose to the spinal cord by 1Gy compared to the anterior model, p=0.001. All template beam plans met OAR criteria, with 1Gy increases noted in maximum heart dose for the 9-field plans, p=0.04. Conclusion: A RapidPlan knowledge-based model for lung SBRT produces comparable results to clinical plans, with increased consistency and greater efficiency. The model encompasses both IMRT and VMAT techniques, differing tumor locations, and beam arrangements. Research supported in part by a grant from Varian Medical Systems, Palo Alto CA.

60 APPLIED LIFE SCIENCES↗

MODEL AND ALGORITHM EVALUATION FOR THE HYBRID UF6 CONTAINER INSPECTION SYSTEM

ABSTRACT Pacific Northwest National Laboratory (PNNL) is developing an automated UF6 cylinder verification station concept based on the combined collection of traditional enrichment-meter (186 keV photons from U-235) data and non-traditional, neutron-induced, high-energy gamma-signatures (3-8.5 MeV) with an array of collimated, medium-resolution scintillators. Previous (2010) work at PNNL demonstrated proof-of-principle that this hybrid method yields accurate, full-volume assay of the cylinder enrichment, reduces systematic errors when compared to several other enrichment assay methods, and provides simplified instrumentation and algorithms suitable for long-term unattended operations. We used Monte Carlo modeling with MCNP5 to support system design (e.g., number and configuration of detector arrays, and design of iron/poly collimators for enhanced (n,γ) conversion) and enrichment algorithm development. We developed a first-generation modeling framework in 2010. These tools have since been expanded, refined and benchmarked against field measurements with a prototype system of a 30B cylinder population (0.2 to 4.95 weight % U-235). The MCNP5 model decomposes the radiation transport problem into a linear superposition of “basis spectra” representing contributions from the different uranium isotopes and gamma-ray generation mechanisms (e.g. neutron capture). This scheme accommodates fast generation of “virtual assay signatures” for arbitrary enrichment, material age, and fill variations. Ongoing (FY-2011) refinements to the physics model include accounting for generation of bremsstrahlung photons, arising primarily from the beta decay of Pa-234m, a U-238 daughter. We are using the refined model to optimize collimator design for the hybrid method. The traditional assay method benefits from a high degree of collimation (to isolate each detector’s field-of-view) and relatively small detector area, while the non-traditional method benefits from a wide field-of-view, i.e. less collimation and larger detectors. We implement the enrichment-meter method by applying a square-wave digital filter to a raw spectrum and extracting the 186-keV peak area directly from the convolute spectrum. Ongoing enhancements to this approach include mitigating a systematic peak-area measurement deficit arising from curvature in the spectrum continuum shape. An optimized system prototype based on model results is utilized in a new set of 2011 field measurements, and model and measurement enrichment assay uncertainties are compared.

MODEL AND ALGORITHM EVALUATION FOR THE HYBRID UF6 ↗

Thresholding Analysis and Feature Extraction from 3D Ground Penetrating Radar Data for Noninvasive Assessment of Peanut Yield

This study explores the efficacy of utilizing a novel ground penetrating radar (GPR) acquisition platform and data analysis methods to quantify peanut yield for breeding selection, agronomic research, and producer management and harvest applications. Sixty plots comprising different peanut market types were scanned with a multichannel, air-launched GPR antenna. Image thresholding analysis was performed on 3D GPR data from four of the channels to extract features that were correlated to peanut yield with the objective of developing a noninvasive high-throughput peanut phenotyping and yield-monitoring methodology. Plot-level GPR data were summarized using mean, standard deviation, sum, and the number of nonzero values (counts) below or above different percentile threshold values. Best results were obtained for data below the percentile threshold for mean, standard deviation and sum. Data both below and above the percentile threshold generated good correlations for count. Correlating individual GPR features to yield generated correlations of up to 39% explained variability, while combining GPR features in multiple linear regression models generated up to 51% explained variability. The correlations increased when regression models were developed separately for each peanut type. This research demonstrates that a systematic search of thresholding range, analysis window size, and data summary statistics is necessary for successful application of this type of analysis. The results also establish that thresholding analysis of GPR data is an appropriate methodology for noninvasive assessment of peanut yield, which could be further developed for high-throughput phenotyping and yield-monitoring, adding a new sensor and new capabilities to the growing set of digital agriculture technologies.

54 ENVIRONMENTAL SCIENCES↗

Stochastic representation and conditioning of process-based geological model by deep generative and recognition networks

Accurate and realistic geological modeling is the core of oil and gas development and production. In recent years, process-based methods are developed to produce highly realistic geological models by simulating the physical processes that reproduce the sedimentary events and develop the geometry. However, the complex dynamic processes are extremely expensive to simulate, making process-based models difficult to be conditioned to field data. In this work, we propose a comprehensive generative adversarial network framework as a machine-learning-assisted approach for mimicking the outputs of process-based geological models with fast generation. The main objective of our work is to obtain a continuous parametrization of the highly realistic process-based geological models which enables us to calibrate the models and condition the models to data. Numerical results are presented to illustrate the capability of our proposed methodology.

58 GEOSCIENCES↗

THE COLLISIONAL EVOLUTION OF DEBRIS DISKS

We explore the collisional decay of disk mass and infrared emission in debris disks. With models, we show that the rate of the decay varies throughout the evolution of the disks, increasing its rate up to a certain point, which is followed by a leveling off to a slower value. The total disk mass falls off {proportional_to}t {sup -0.35} at its fastest point (where t is time) for our reference model, while the dust mass and its proxy-the infrared excess emission-fades significantly faster ({proportional_to}t {sup -0.8}). These later level off to a decay rate of M{sub tot}(t){proportional_to}t {sup -0.08} and M{sub dust}(t) or L{sub ir}(t){proportional_to}t {sup -0.6}. This is slower than the {proportional_to}t {sup -1} decay given for all three system parameters by traditional analytic models. We also compile an extensive catalog of Spitzer and Herschel 24, 70, and 100 {mu}m observations. Assuming a log-normal distribution of initial disk masses, we generate model population decay curves for the fraction of stars harboring debris disks detected at 24 {mu}m. We also model the distribution of measured excesses at the far-IR wavelengths (70-100 {mu}m) at certain age regimes. We show general agreement at 24 {mu}m between the decay of our numerical collisional population synthesis model and observations up to a Gyr. We associate offsets above a Gyr to stochastic events in a few select systems. We cannot fit the decay in the far-infrared convincingly with grain strength properties appropriate for silicates, but those of water ice give fits more consistent with the observations (other relatively weak grain materials would presumably also be successful). The oldest disks have a higher incidence of large excesses than predicted by the model; again, a plausible explanation is very late phases of high dynamical activity around a small number of stars. Finally, we constrain the variables of our numerical model by comparing the evolutionary trends generated from the exploration of the full parameter space to observations. Amongst other results, we show that erosive collisions are dominant in setting the timescale of the evolution and that planetesimals on the order of 100 km in diameter are necessary in the cascades for our population synthesis models to reproduce the observations.

79 ASTRONOMY AND ASTROPHYSICS↗

Gearbox and Drivetrain Models to Study Dynamic Effects of Modern Wind Turbines: Preprint

Wind turbine drivetrains consist of components that directly convert kinetic energy from the wind to electrical energy. Guaranteeing robust and reliable drivetrain designs is therefore important to minimize turbine downtime. Current drivetrain models often lack the ability to model both the impacts of electrical transients as well as wind turbulence and shear in one package. In this work, thecapability of the FAST wind turbine computer-aided engineering tool, developed by the National Renewable Energy Laboratory, is enhanced through integration of a dynamic model of the drivetrain. The dynamic drivetrain model is built using Simscape in the MATLAB/Simulink environment and incorporates detailed electrical generator models. This model can be used in the future to test advanced controlschemes to extend life of the gearbox.

17 WIND ENERGY↗