Search NASA⌕ Search

SEARCH · Search NASA

Results for “gradient methods”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,207 records · Page 67

Methods, software and datasets to verify DVH calculations against analytical values: Twenty years late(r)

Purpose: The authors designed data, methods, and metrics that can serve as a standard, independent of any software package, to evaluate dose-volume histogram (DVH) calculation accuracy and detect limitations. The authors use simple geometrical objects at different orientations combined with dose grids of varying spatial resolution with linear 1D dose gradients; when combined, ground truth DVH curves can be calculated analytically in closed form to serve as the absolute standards. Methods: DICOM RT structure sets containing a small sphere, cylinder, and cone were created programmatically with axial plane spacing varying from 0.2 to 3 mm. Cylinders and cones were modeled in two different orientations with respect to the IEC 1217 Y axis. The contours were designed to stringently but methodically test voxelation methods required for DVH. Synthetic RT dose files were generated with 1D linear dose gradient and with grid resolution varying from 0.4 to 3 mm. Two commercial DVH algorithms—PINNACLE (Philips Radiation Oncology Systems) and PlanIQ (Sun Nuclear Corp.)—were tested against analytical values using custom, noncommercial analysis software. In Test 1, axial contour spacing was constant at 0.2 mm while dose grid resolution varied. In Tests 2 and 3, the dose grid resolution was matched to varying subsampled axial contours with spacing of 1, 2, and 3 mm, and difference analysis and metrics were employed: (1) histograms of the accuracy of various DVH parameters (total volume, D{sub max}, D{sub min}, and doses to % volume: D99, D95, D5, D1, D0.03 cm{sup 3}) and (2) volume errors extracted along the DVH curves were generated and summarized in tabular and graphical forms. Results: In Test 1, PINNACLE produced 52 deviations (15%) while PlanIQ produced 5 (1.5%). In Test 2, PINNACLE and PlanIQ differed from analytical by >3% in 93 (36%) and 18 (7%) times, respectively. Excluding D{sub min} and D{sub max} as least clinically relevant would result in 32 (15%) vs 5 (2%) scored deviations for PINNACLE vs PlanIQ in Test 1, while Test 2 would yield 53 (25%) vs 17 (8%). In Test 3, statistical analyses of volume errors extracted continuously along the curves show PINNACLE to have more errors and higher variability (relative to PlanIQ), primarily due to PINNACLE’s lack of sufficient 3D grid supersampling. Another major driver for PINNACLE errors is an inconsistency in implementation of the “end-capping”; the additional volume resulting from expanding superior and inferior contours halfway to the next slice is included in the total volume calculation, but dose voxels in this expanded volume are excluded from the DVH. PlanIQ had fewer deviations, and most were associated with a rotated cylinder modeled by rectangular axial contours; for coarser axial spacing, the limited number of cross-sectional rectangles hinders the ability to render the true structure volume. Conclusions: The method is applicable to any DVH-calculating software capable of importing DICOM RT structure set and dose objects (the authors’ examples are available for download). It includes a collection of tests that probe the design of the DVH algorithm, measure its accuracy, and identify failure modes. Merits and applicability of each test are discussed.

60 APPLIED LIFE SCIENCES↗

Scalable and Energy-Efficient Methods for Interactive Exploration of Scientific Data

The main scientific contributions of this project are the following novel concepts for multidimensional arrays: shape-based similarity join (SIGMOD 2016), incremental view maintenance (SIGMOD 2017), user-defined stencil functions (HPDC 2017), and distributed caching for in-situ processing (SSDBM 2018). Building on our collaboration with the astrophysics group at LBNL, we applied these techniques to the data generated in the Palomar Transient Factory (PTF) astronomical survey. They played a pivotal role in the first-ever observation of a neutron star merger, which produces gravitational waves and turns out to be the origin of heavy elements, including gold. This has lead to a Science magazine article that has received extensive media coverage on ACM TechNews, Slashdot, FiveThirtyEight, and Quanta Magazine, among others. Additionally, two other articles detailing related aspects of the same discovery have been published in the Astrophysical Journal Letters journal. These publications have more than 3,000 citations according to Google Scholar (as of February 2022). This cross-disciplinary collaboration provided very good opportunities to apply database techniques to real-life scientific problems. The fact that they facilitated major discoveries in astrophysics proves the importance of our research. In addition to the work on multidimensional array databases, this project has also developed stochastic gradient descent (SGD) optimization algorithms for training large scale machine learning models, methods for querying in-situ data, and a database query optimizer based on sketch synopses.

79 ASTRONOMY AND ASTROPHYSICS↗

Phonon thermal transport outside of local equilibrium in nanowires via molecular dynamics

We study thermal transport through Pt nanowires that bridge planar contacts as a function of wire length and vibrational frequency of the contacts. When phonons in the contacts have lower average frequencies than those in the wires thermal transport occurs under conditions away from local equilibrium with low-frequency phonons experiencing a higher thermal gradient than high-frequency ones. This results in a size-dependent increase in the effective thermal conductivity of the wire with decreasing vibrational frequencies of the contacts. The interfacial resistivity when heat flows from the wire to the contact is also size-dependent and has the same physical origin in the lack of full equilibration in short nanowires. We develop a model based on a 1D atomic chain that captures the salient physics of the MD results.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Thermometry of an open spatially inhomogeneous flame by diode laser absorption spectroscopy

Features of a single-beam variant of diode laser absorption spectroscopy (DLAS) used for temperature measurements in a spatially inhomogeneous hot zone are studied. A flame temperature of a commercial burner is measured by using water molecule absorption lines. The flame is probed by a diode laser radiation at various heights and in various directions. The correctness of DLAS measurement results is verified by comparing DLAS data with temperatures measured by the method of coherent anti-Stokes scattering (CARS). It is shown that in homogeneous flame domains the temperatures obtained by both the methods coincide. Temperature inhomogeneity along the probe beam direction is a principal problem of DLAS, in which case the method yields information about some effective temperature along the light trace. The difference in data obtained by DLAS and CARS is estimated. In probing traces with temperature gradients the difference reaches 700 K. Possibilities of accounting for temperature inhomogeneities along the trace in DLAS measurements are discussed. A method of differential spectra is suggested, which in certain conditions can provide substantially better coincidence of results for the two methods. (laser spectroscopy)

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

THE METALLICITY GRADIENT OF THE THICK DISK BASED ON RED HORIZONTAL-BRANCH STARS FROM SDSS DR8

Based on SDSS DR8, we have selected a sample of 1728 red horizontal-branch (RHB) stars with |Z| < 3 kpc by using a color-metallicity relation and stellar parameters. The sample stars clearly trace a typical thick disk population with peaks at |Z| = 1.26 kpc and [Fe/H] = -0.54. The vertical metallicity gradient of the thick disk is estimated in two ways. One is a fit to the Gaussian peaks of the metallicity histograms of the thick disk by subtracting minor contributions from the thin disk and the inner halo based on the Besancon Galaxy Model. The resulting gradient is -0.12 {+-} 0.01 dex kpc{sup -1} for 0.5 < |Z| < 3 kpc. The other method is to linearly fit the data based on stars with 1 < |Z| < 3 kpc being the main component of the thick disk. Five subgroups are then selected in different directions in the X-|Z| plane to investigate the difference in the vertical metallicity gradient between the Galactocenter and anti-Galactocenter directions. We found that a vertical gradient of -0.22 {+-} 0.07 dex kpc{sup -1} is detected for five directions except for one involving the pollution of stars from the bulge. The results indicate that the vertical gradient is dominant, but a radial gradient has a minor contribution for the thick disk population represented by RHB stars with 1 < |Z| < 3 kpc. The present work strongly suggests the existence of a metallicity gradient in the thick disk, which is thought to be negligible in most previous works in the literature.

79 ASTRONOMY AND ASTROPHYSICS↗

Theoretical Investigation of Structural, Electronic, Magnetic, and Mechanical Properties of Quaternary Heusler Alloys CoVTiX (X = As, Si)

The structural, electronic, magnetic, and mechanical properties of CoVTiAs and CoVTiSi quaternary Heusler alloys are studied using the full-potential linearized augmented plane-wave (FP-LAPW) method in the framework of the density functional theory (DFT). The generalized gradient approximation (GGA) is chosen for the exchange-correlation energy, whereas the modified Becke-Johnson (mBJ) formalism is applied for the electronic properties. Both CoVTiAs and CoVTiSi are stable in the type 2 (FM) structure. The mechanical properties of CoVTiX (X = As, Si) are predicted from the calculated elastic constants. Results identify that these compounds are mechanically stable. Our results with the mBJ approximation predict that CoVTiAs and CoVTiSi alloys are half-metallic ferromagnets with band gaps of 0.450 and 0.424 eV, respectively.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Portable dual field gradient force multichannel flow cytometer device with a dual wavelength low noise detection scheme

Systems and methods for combining dielectrophoresis, magnetic forces, and hydrodynamic forces to manipulate particles in channels formed on top of an electrode substrate are discussed. A magnet placed in contact under the electrode substrate while particles are flowing within the channel above the electrode substrate allows these three forces to be balanced when the system is in operation. An optical detection scheme using near-confocal microscopy for simultaneously detecting two wavelengths of light emitted from the flowing particles is also discussed.

47 OTHER INSTRUMENTATION↗

HDBind: encoding of molecular structure with hyperdimensional binary representations

Traditional methods for identifying “hit” molecules from a large collection of potential drug-like candidates rely on biophysical theory to compute approximations to the Gibbs free energy of the binding interaction between the drug and its protein target. These approaches have a significant limitation in that they require exceptional computing capabilities for even relatively small collections of molecules. Increasingly large and complex state-of-the-art deep learning approaches have gained popularity with the promise to improve the productivity of drug design, notorious for its numerous failures. However, as deep learning models increase in their size and complexity, their acceleration at the hardware level becomes more challenging. Hyperdimensional Computing (HDC) has recently gained attention in the computer hardware community due to its algorithmic simplicity relative to deep learning approaches. The HDC learning paradigm, which represents data with high-dimension binary vectors, allows the use of low-precision binary vector arithmetic to create models of the data that can be learned without the need for the gradient-based optimization required in many conventional machine learning and deep learning methods. This algorithmic simplicity allows for acceleration in hardware that has been previously demonstrated in a range of application areas (computer vision, bioinformatics, mass spectrometery, remote sensing, edge devices, etc.). To the best of our knowledge, our work is the first to consider HDC for the task of fast and efficient screening of modern drug-like compound libraries. We also propose the first HDC graph-based encoding methods for molecular data, demonstrating consistent and substantial improvement over previous work. We compare our approaches to alternative approaches on the well-studied MoleculeNet dataset and the recently proposed LIT-PCBA dataset derived from high quality PubChem assays. We demonstrate our methods on multiple target hardware platforms, including Graphics Processing Units (GPUs) and Field Programmable Gate Arrays (FPGAs), showing at least an order of magnitude improvement in energy efficiency versus even our smallest neural network baseline model with a single hidden layer. Our work thus motivates further investigation into molecular representation learning to develop ultra-efficient pre-screening tools. We make our code publicly available at https://github.com/LLNL/hdbind.

59 BASIC BIOLOGICAL SCIENCES↗

Toward Establishing Uniqueness of Experimentally Determined Transference Numbers

The passage of current through a battery results in the development of concentration gradients in the electrolytic phase. For a fully characterized binary electrolyte, where the conductivity, salt diffusion coefficient, cation transference number, and the thermodynamic factor are known, concentration and potential gradients in the electrolytic phase can be modeled using Newman’s concentrated solution theory. We report two methods for measuring the transference number: the standard method based on electrochemical measurements ( t + , echem 0 ) and electrophoretic NMR ( t + , eNMR 0 ). The electrochemical approach requires combining measurements from multiple experiments; the equations used to determine the cation transference number and the thermodynamic factor are coupled, nonlinear algebraic equations. In the electrophoretic-NMR-based approach, however, the equations used to determine the cation transference number and the thermodynamic factor are decoupled. We find for a liquid electrolyte comprised of a lithium salt dissolved in tetraglyme, the values of the transference numbers obtained by these two methods are distinct. For example, at 30 °C, t + , echem 0 = −1.02 ± 1.11 and t + , eNMR 0 = 0.25 ± 0.04. The corresponding thermodynamic factors are also different. While the magnitude of the predicted concentration gradients based on the two sets of parameters are different, the predicted current-voltage relationships are similar.

Hickson, Darby T. (ORCID:0000000251339755)↗

Seismic Spatial Gradients and Machine Learning-Based Classifiers for Explosion Monitoring (LDRD 218327)

This final report summarizes the work completed under the Laboratory Directed Research and Development (LDRD) project “Seismic Spatial Gradients as a Machine Learning-Based Classifier for Explosion Monitoring.” The overarching goal of the project was to explore the efficacy of using machine learning-based classification algorithms where the input data are the spatial gradient of the seismic wavefield collected at a single point on the Earth’s surface. The methods that I describe here are in direct contrast to conventional methods of seismic discrimination which typically rely on a spatially extended network of instruments and physics-based wavefield attributes such as, for example, the ratio between $\textit{P}$ and $\textit{S}$ waves. Rather, we use the spatial gradient of the seismic wavefield observed at a single point on the Earth’s surface and data processing approaches inspired by the machine learning community. We tested two algorithms, a neural network and a modified version of principal component analysis termed Spectrally Filtered Principal Component Analysis (SFPCA). To test these algorithms, we first conducted a series of numerical tests using synthetic data and then conducted a small-scale controlled field experiment. The tests using synthetic data showed that both algorithms had high success rates on gradiometric data, even when simulated noise was added to the signal. Furthermore, we found that using seismic spatial gradients increased the performance of our discrimination algorithms when compared to using just the traditional translational motion seismic data. The tests with field data also showed a high degree of discriminative success.

58 GEOSCIENCES↗

Sequential Kalman tuning of the t -preconditioned Crank-Nicolson algorithm: efficient, adaptive and gradient-free inference for Bayesian inverse problems

Ensemble Kalman Inversion (EKI) has been proposed as an efficient method for the approximate solution of Bayesian inverse problems with expensive forward models. However, when applied to the Bayesian inverse problem EKI is only exact in the regime of Gaussian target measures and linear forward models. Here, in this work we propose embedding EKI and Flow Annealed Kalman Inversion, its normalizing flow (NF) preconditioned variant, within a Bayesian annealing scheme as part of an adaptive implementation of the t-preconditioned Crank-Nicolson (tpCN) sampler. The tpCN sampler differs from standard pCN in that its proposal is reversible with respect to the multivariate t-distribution. The more flexible tail behaviour allows for better adaptation to sampling from non-Gaussian targets. Within our Sequential Kalman Tuning (SKT) adaptation scheme, EKI is used to initialize and precondition the tpCN sampler for each annealed target. The subsequent tpCN iterations ensure particles are correctly distributed according to each annealed target, avoiding the accumulation of errors that would otherwise impact EKI. We demonstrate the performance of SKT for tpCN on three challenging numerical benchmarks, showing significant improvements in the rate of convergence compared to adaptation within standard SMC with importance weighted resampling at each temperature level, and compared to similar adaptive implementations of standard pCN. The SKT scheme applied to tpCN offers an efficient, practical solution for solving the Bayesian inverse problem when gradients of the forward model are not available. Code implementing the SKT schemes for tpCN is available at https://github.com/RichardGrumitt/KalmanMC.

97 MATHEMATICS AND COMPUTING↗

Defect And Damage Characterization Of Additively Manufactured Titanium Alloy Ti-5553 Using Traditional Computed Tomography Volume Segmentation And Machine Learning Algorithms

The mechanical response of a component is affected by defects, such as porosity, arising from the laser powder bed fusion (LPBF) fabrication process. Thus, it is important to develop accurate and efficient inspection methods for identifying porosity. In this work, porosity identified in an X-ray computed tomography (XCT) volume of a Ti-5553 coupon was compared to pores identified in a serial sectioned volume that represented the ground truth. The porosity of the XCT scan was identified using contrast-based, ISO-based, and machine learning (ML) methods for segmentation. Large inherent porosity was easy to identify, but the ISO thresholding still struggled due to the intensity gradient resulting from both the beam hardening in XCT and the uneven lighting of the serial sectioning panels. Further, the results show that ML-based methods were better suited for identifying small pores and reducing the amount of false positives. Additionally, high strain-rate impact testing was done on some of the XCT samples as well as post-mortem XCT inspection, and the same suite of segmentation and quantification tools were used to identify the large spallation cavities. The comparison of porosity pre- and post-mortem provides insight on the influence of the LPBF porosity on the formation of spall cavities.

36 MATERIALS SCIENCE↗

Comprehensive assessment of deep reinforcement learning approaches for economic dispatch in nuclear-driven microgrids

As the electrical grid integrates more variable renewable energy sources such as wind and solar, the demand for distributed and flexible systems to address this increased variability becomes critical. Nuclear-driven microgrids provide a promising solution by offering stable generation to complement intermittent renewables, ensuring grid reliability and operating efficiency. This paper proposes a recurrent deep reinforcement learning framework for optimal economic dispatch in a nuclear-powered microgrid integrating renewable energy sources, small modular reactors, battery storage systems, and balance-of-plant dynamics. A three-agent control architecture is developed, where demand and renewable energy agents act as forecasters, and a reinforcement learning-based dispatch agent performs real-time energy allocation. A nonlinear programming formulation is first used to generate an optimal baseline for benchmarking. The proposed dispatch controller, based on Proximal Policy Optimization enhanced with Long Short-Term Memory networks, exploits temporal correlations in system dynamics by taking advantage of the time series used as inputs to improve policy robustness under uncertainty. Comparative analysis against established deep reinforcement learning methods, including Proximal Policy Optimization with a feedforward architecture, Soft Actor-Critic, and Twin Delayed Deep Deterministic Policy Gradient, demonstrates superior performance. Numerical results indicate that the proposed controller achieves a 0.39% cost reduction relative to the nonlinear programming benchmark and outperforms other learning-based methods by generating additional revenue of up to 0.35%. All reinforcement learning controllers compute dispatch actions in less than 0.3 s, resulting in a computational speedup of more than three orders of magnitude over the nonlinear programming baseline. The findings of this paper highlight their applicability for real-time operation and control in nuclear-integrated microgrids under volatile operating conditions.

24 POWER TRANSMISSION AND DISTRIBUTION↗

CCSI Toolset 3.19 Release

CCSI Toolset 3.19 Release Highlights A gradient generation tool was developed to support GENN models in FOQUS. Certain machine learning tools train gradient-enhanced neural network (GENN) models which can be more accurate for complex datasets given a priori knowledge of model derivatives. However, the derivatives must be known beforehand and are not often available for process data. This tool automatically predicts the gradients for a training dataset in a form usable by common GENN trainers, such as Surrogate Modeling Toolbox. Support was added for Surrogate Modeling Toolbox GENN models in FOQUS, including updates to the run methods, node properties, test framework, documentation and optional dependencies list. Users can train/save Surrogate Modeling Toolbox gradient-enhanced neural network (GENN) models with custom objects and produce .pkl files compatible with the Machine Learning/Artificial Intelligence Plugin in FOQUS. A simpler implementation of the ordering algorithm in the Sequential Design of Experiments (SDOE) module was included. The SDOE examples documentation was updated. The Optimality-Based Design of Experiments was updated to improve the error handling when the results are None.

AS↗

Multi-Agent Safe Policy Learning for Power Management of Networked Microgrids

This paper presents a supervised multi-agent safe policy learning (SMAS-PL) method for optimal power management of networked microgrids (MGs) in distribution systems. While unconstrained reinforcement learning (RL) algorithms are black-box decision models that could fail to satisfy grid operational constraints, our proposed method considers AC power flow equations and other operational limits. Accordingly, the training process employs the gradient information of operational constraints to ensure that the optimal control policy functions generate safe and feasible decisions. Furthermore, we have developed a distributed consensus-based optimization approach to train the agents’ policy functions while maintaining MGs’ privacy and data ownership boundaries. After training, the learned optimal policy functions can be safely used by the MGs to dispatch their local resources, without the need to solve a complex optimization problem from scratch. Lastly, numerical experiments have been devised to verify the performance of the proposed method.

, policy gradient↗

Integrated Model, Batch and Domain Parallelism in Training Neural Networks

We propose a new integrated method of exploiting model, batch and domain parallelism for the training of deep neural networks (DNNs) on large distributed-memory computers using minibatch stochastic gradient descent (SGD). Our goal is to find an efficient parallelization strategy for a fixed batch size using $P$ processes. Our method is inspired by the communication-avoiding algorithms in numerical linear algebra. We see $P$ processes as logically divided into a $P_r \times P_c$ grid where the $P_r$ dimension is implicitly responsible for model/domain parallelism and the $P_c$ dimension is implicitly responsible for batch parallelism. In practice, the integrated matrix-based parallel algorithm encapsulates these types of parallelism automatically. We analyze the communication complexity and analytically demonstrate that the lowest communication costs are often achieved neither with pure model nor with pure data parallelism. We also show how the domain parallel approach can help in extending the theoretical scaling limit of the typical batch parallel method.

97 MATHEMATICS AND COMPUTING↗

X-band EPR imaging as a tool for gradient dose reconstruction in irradiated bones

Purpose: Various tools are currently available for dose reconstruction in individuals after accidental exposure to ionizing radiation. Among the available biological analyses, Monte Carlo simulations, and biophysical methods, such as electron paramagnetic resonance (EPR), the latter has proved its usefulness for retrospective dosimetry. Although EPR spectroscopy is probably the most sensitive technique, it does not provide spatial dosimetric data. This information is, however, highly desirable when steep dose gradient irradiations are involved. The purpose of this work was to explore the possibilities of EPR imaging (EPRI) for spatial dose reconstruction in irradiated biological material. Methods: X-band EPRI was used to reconstruct ex vivo the relative dose distribution in human bone samples and hydroxyapatite phantoms after irradiation with brachytherapy seeds or x rays. Three situations were investigated: Homogeneous, stepwise gradient, and continuous gradient irradiation. Results: EPRI gave a faithful relative spin density distribution in bone samples and in hydroxyapatite phantoms. Measured dose ratios were in close agreement with the actual delivered dose ratios. EPRI was able to distinguish the dose gradients induced by two different sources ({sup 125}I and {sup 192}Ir). However, the measured spatial resolution of the system was 1.9 mm and this appeared to be a limiting factor. The method could be improved by using new signal postprocessing strategies. Conclusions: This study demonstrates that EPRI can be used to assess the regional relative dose distribution in irradiated bone samples. The method is currently applicable to ex vivo measurements of small size samples with low variation in tissue density but is likely to be adapted for in vivo application using L-band EPRI.

61 RADIATION PROTECTION AND DOSIMETRY↗

Friction-Induced Fluid Heating in Nanoscale Helium Flows

We investigate the mechanism of friction-induced fluid heating in nanoconfinements. Molecular dynamics simulations are used to study the temperature variations of liquid helium in nanoscale Poiseuille flows. It is found that the fluid heating is dominated by different sources of friction as the external driving force is changed. For small external force, the fluid heating is mainly caused by the internal viscous friction in the fluid. When the external force is large and causes fluid slip at the surfaces of channel walls, the friction at the fluid-solid interface dominates over the internal friction in the fluid and is the major contribution to fluid heating. An asymmetric temperature gradient in the fluid is developed in the case of nonidentical walls and the general temperature gradient may change sign as the dominant heating factor changes from internal to interfacial friction with increasing external force.

77 NANOSCIENCE AND NANOTECHNOLOGY↗