Search NASA⌕ Search

SEARCH · Search NASA

Results for “gradient calculation”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2017

Analytical gradients for tensor hyper-contracted MP2 and SOS-MP2 on graphical processing units

Analytic energy gradients for tensor hyper-contraction (THC) are derived and implemented for second-order Møller-Plesset perturbation theory (MP2), with and without the scaled-opposite-spin (SOS)-MP2 approximation. By exploiting the THC factorization, the formal scaling of MP2 and SOS-MP2 gradient calculations with respect to system size is reduced to quartic and cubic, respectively. An efficient implementation has been developed that utilizes both graphics processing units and sparse tensor techniques exploiting spatial sparsity of the atomic orbitals. THC-MP2 has been applied to both geometry optimization and ab initio molecular dynamics (AIMD) simulations. Furthermore, the resulting energy conservation in micro-canonical AIMD demonstrates that the implementation provides accurate nuclear gradients with respect to the THC-MP2 potential energy surfaces.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2019

Improving efficiency of semi‐direct møller–plesset second‐order perturbation methods through oversubscription on multiple nodes

The purpose of this work is to evaluate the efficacy of oversubscription, at the 1 n , 2 n , and 3 n levels for n physical cores, on semi‐direct MP2 methods within NWChem when using two and three Intel nodes. Semi‐direct MP2 energy and gradient calculations were performed on chemical systems ranging from 824 to 1626 basis functions using the cc‐pVDZ basis set. Wall times for semi‐direct MP2 energies were reduced by as much as 36% using two nodes and 44% using three nodes compared to no oversubscription. Total energy consumed by the CPU and DRAM was also reduced by as much as 12% using two nodes and as much as 20% using three nodes when oversubscribing. MP2 gradient wall times improved by as much as 16% using two nodes and 18% using three nodes compared to execution at the 1 n level; however, energy savings were insignificant. Intel performance‐counter data show a strong correlation between total wall time saved and less time spent in the idle state, indicating a more efficient use of the processors when oversubscribing. © 2019 Wiley Periodicals, Inc.

Fought, Ellie L.↗
DOE OSTI2022

A scalable compressible volume of fluid solver using a stratified flow model

Here, in this paper, a Volume of Fluid method which uses a Stratified Flow model for flux calculation is proposed to solve a compressible multiphase flow in a way that has high parallel efficiency. The solver is part of an open-source computing tool to study fuel entrainment and combustion in the application of paraffin-based hybrid rocket motors. This paper focuses on the compressible Volume of Fluid solver for the liquid and gaseous phases in this environment. This solver uses perfect gas and stiffened gas models for equation of state, with the ability to easily add other models as desired. Additionally a comparison of a Riemann solver versus an AUSM+up scheme is conducted, showing improvement with the later in the solution of multiphase flows. This solver successfully reproduces shock tube cases with the expected accuracy, as well as simulating two-dimensional shear and gravity-driven flows. Results also show that the code is able to utilize a Stratified Flow model to evaluate shear flow without the need for interface reconstruction or gradient calculations. Finally, it is shown that the solver has near-ideal scaling under strong scaling tests as well as good performance in static scaling, giving improved performance over current options and implying future advancement of high-performance computing efficiency for multiphase flow solvers.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Sparsity for Gradient-Based Optimization of Wind Farm Layouts

Optimizing wind farm layouts is an important step in designing an efficient wind farm. Optimizing wind farm layouts is also a difficult task due to computation times increasing with the number of turbines present in the farm. The most computationally expensive part of gradient-based optimization is calculating the gradient. In order to reduce the expense of gradient calculation, we performed a study on the use of sparsity in wind farm layout optimization. This paper presents the findings of the sparsity study and provides a method to use sparsity in wind farm layout optimization. We tested this sparsity method by optimizing multiple farms with sparse methods and compared the results to optimizations of the same farms using traditional methods. By using the sparse method to optimize multiple farms, we found that the objective results were comparable between sparse and traditional methods and that sparse methods were 4 times faster than traditional methods on average. We expect more speedups with improved methodology and larger wind farms. By using sparse methods, it is possible to solve the wind farm layout optimization problem more efficiently, thus allowing for a more thorough study of the wind farm layout design space without excessive computational costs. Further work is required to refine the method and prepare for testing on real-world wind farm layout applications.

gradient↗
DOE OSTI2016

Three-dimensional local residual stress and orientation gradients near graphite nodules in ductile cast iron [3D local residual stress and orientation gradients near graphite nodules in ductile cast iron]

A synchrotron technique, differential aperture X-ray microscopy (DAXM), has been applied to characterize the microstructure and analyze the local mesoscale residual elastic strain fields around graphite nodules embedded in ferrite matrix grains in ductile cast iron. Compressive residual elastic strains are measured with a maximum strain of ~6.5–8 × 10 –4 near the graphite nodules extending into the matrix about 20 μm, where the elastic strain is near zero. The experimental data are compared with a strain gradient calculated by a finite element model, and good accord has been found but with a significant overprediction of the maximum strain. This is discussed in terms of stress relaxation during cooling or during storage by plastic deformation of the nodule, the matrix or both. Furthermore, relaxation by plastic deformation of the ferrite is demonstrated by the formation of low energy dislocation cell structure also quantified by the DAXM technique.

36 MATERIALS SCIENCE↗
DOE OSTI2021

On the feasibility of using physics-informed machine learning for underground reservoir pressure management

In this work, we evaluate the feasibility of using physics-informed machine learning (PIML) for underground energy-related pressure management. To this end, we develop a PIML framework to manage underground reservoir pressures by training neural networks to determine fluid extraction rates for dedicated extraction wells during fluid injection operations given a range of reservoir conditions (e.g., transmissivity and storativity). We implement an automatically-differentiable analytical physics model of fluid flow in porous media within the PIML framework as a proxy for more complicated models. This allows us to execute a sufficient number of training scenarios to fully evaluate the feasibility of using PIML to support pressure management activities. We quantify the number of physics-model parameters required for automatic differentiation to become more efficient than finite-difference gradient calculations. We use a simple scenario with a single injector, extractor, and critical location for our feasibility analysis. We evaluate the effect of the size of the training dataset (i.e., the number of reservoir condition samples) on the accuracy and efficiency of the PIML framework. For an equivalent number of model evaluations, the larger training dataset took less time to train and produced a neural network that was able to more accurately manage reservoir pressures. We also evaluate the effect of the training dataset batch size (i.e., number of reservoir condition samples used to update the neural network coefficients during training; i.e., how the training dataset is partitioned). While training ran faster with larger batch sizes, they produced neural networks that managed pressures less accurately. We demonstrate the approach on a more complex scenario involving 10 injectors, 10 extractors, and 4 critical locations (a relatively high well density of 20 wells/km2). We provide the number of forward and adjoint model evaluations required in each case as an indication of the feasibility of using PIML for pressure management when more complicated physics models with longer execution times are used.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2015

Second order finite volume scheme for Maxwell's equations with discontinuous electromagnetic properties on unstructured meshes

This paper presents a second order finite volume scheme for numerical solution of Maxwell's equations with discontinuous dielectric permittivity and magnetic permeability on unstructured meshes. The scheme is based on Godunov scheme and employs approaches of Van Leer and Lax–Wendroff to increase the order of approximation. To keep the second order of approximation near dielectric permittivity and magnetic permeability discontinuities a novel technique for gradient calculation and limitation is applied near discontinuities. Results of test computations for problems with linear and curvilinear discontinuities confirm second order of approximation. The scheme was applied to modelling propagation of electromagnetic waves inside photonic crystal waveguides with a bend.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2023

Differentiable modelling to unify machine learning and physical models for geosciences

Process-based modelling offers interpretability and physical consistency in many domains of geosciences but struggles to leverage large datasets efficiently. Machine-learning methods, especially deep networks, have strong predictive skills yet are unable to answer specific scientific questions. Here, in this Perspective, we explore differentiable modelling as a pathway to dissolve the perceived barrier between process-based modelling and machine learning in the geosciences and demonstrate its potential with examples from hydrological modelling. ‘Differentiable’ refers to accurately and efficiently calculating gradients with respect to model variables or parameters, enabling the discovery of high-dimensional unknown relationships. Differentiable modelling involves connecting (flexible amounts of) prior physical knowledge to neural networks, pushing the boundary of physics-informed machine learning. It offers better interpretability, generalizability, and extrapolation capabilities than purely data-driven machine learning, achieving a similar level of accuracy while requiring less training data. Additionally, the performance and efficiency of differentiable models scale well with increasing data volumes. Under data-scarce scenarios, differentiable models have outperformed machine-learning models in producing short-term dynamics and decadal-scale trends owing to the imposed physical constraints. Differentiable modelling approaches are primed to enable geoscientists to ask questions, test hypotheses, and discover unrecognized physical relationships. Future work should address computational challenges, reduce uncertainty, and verify the physical significance of outputs.

58 GEOSCIENCES↗
DOE OSTI2021

Stiff neural ordinary differential equations

Neural Ordinary Differential Equations (ODEs) are a promising approach to learn dynamical models from time-series data in science and engineering applications. This work aims at learning neural ODEs for stiff systems, which are usually raised from chemical kinetic modeling in chemical and biological systems. We first show the challenges of learning neural ODEs in the classical stiff ODE systems of Robertson’s problem and propose techniques to mitigate the challenges associated with scale separations in stiff systems. We then present successful demonstrations in stiff systems of Robertson’s problem and an air pollution problem. The demonstrations show that the usage of deep networks with rectified activations, proper scaling of the network outputs as well as loss functions, and stabilized gradient calculations are the key techniques enabling the learning of stiff neural ODEs. The success of learning stiff neural ODEs opens up possibilities of using neural ODEs in applications with widely varying time-scales, such as chemical dynamics in energy conversion, environmental engineering, and life sciences.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Stress field measurements using quantitative schlieren

Quantitative schlieren analysis is extended here to optically transparent solids in quasi-static and dynamic experiments to measure stress distributions. The quasi-static experiments in polymethyl methacrylate (PMMA) compared refraction angles and stress gradients calculated from schlieren images to the analytical Flamant solution of a line load on a half-space. The quantitative schlieren measurements of the stress field in the thin sample with a load compared well to the analytical solution. The analysis method was then extended to explosive induced shock waves in PMMA. The explosive induced response of PMMA was experimentally studied using high-speed schlieren to visualize the shock propagation in conjunction with Photon Doppler Velocimetry (PDV) to record surface velocity histories. The stress state estimated from the schlieren images was compared to the stress calculated from the PDV measurements. High-speed imaging limitations caused the shock wave to not be fully resolved in the images, but was resolved in the PDV measurement. The stress state behind the shock calculated from the high-speed images followed a similar trend to the stress calculated from the PDV measurements.

Physics↗
DOE OSTI2026

Efficient simulation of optical spectra via machine learning and physical decomposition of environmental effects

Simulations of optical spectra can provide key insights to aid experimental interpretation of electronic excitation phenomena. For chromophores in the condensed phase, these spectra, which incorporate the coupling between electronic excitation and molecular and solvent nuclear motions, can be simulated using excitation energies obtained from molecular dynamics simulations of the chromophore and solvent. Here, we present a hybrid scheme that exploits machine learning and physically informed spectral densities to show that as few as 25 ground and excited state energetic gradient calculations can be used to construct models that accurately predict environment-influenced vibronic coupling in optical spectra. We demonstrate our approach for the green fluorescent protein chromophore in water and the cresyl violet chromophore in methanol. We show that our hybrid approach, employing a machine learning model for the high-frequency spectral density and an ab initio parameterized Debye spectral density for the low-frequency, results in a systematic improvement of the optical absorption lineshape, leading to a simple machine learning scheme that can be used for the simulation of spectral densities and optical spectra.

Snider, Andrew [Univ. of California, Merced, CA (U↗
DOE OSTI2015

MOLECULAR GAS AND STAR-FORMATION PROPERTIES IN THE CENTRAL AND BAR REGIONS OF NGC 6946

In this work, we investigate the molecular gas and star-formation properties in the barred spiral galaxy NGC 6946 using multiple molecular lines and star-formation tracers. A high-resolution image (100 pc) of {sup 13}CO (1–0) is created for the inner 2 kpc disk by the single-dish Nobeyama Radio Observatory 45 m telescope and interferometer Combined Array for Research in Millimeter-wave Astronomy, including the central region (nuclear ring and bar) and the offset ridges of the primary bar. Single-dish HCN (1–0) observations were also made to constrain the amount of dense gas. The physical properties of molecular gas are inferred from (1) the large velocity gradient calculations using our observations and archival {sup 12}CO (1–0), {sup 12}CO(2–1) data, (2) the dense gas fraction suggested by the luminosity ratio of HCN to {sup 12}CO (1–0), and (3) the infrared color. The results show that the molecular gas in the central region is warmer and denser than that of the offset ridges. The dense gas fraction of the central region is similar to that of luminous infrared galaxies/ultraluminous infrared galaxies, whereas the offset ridges are close to the global average of normal galaxies. The coolest and least-dense region is found in a spiral-like structure, which was misunderstood to be part of the southern primary bar in previous low-resolution observations. The star-formation efficiency (SFE) changes by about five times in the inner disk. The variation of SFE agrees with the prediction in terms of star formation regulated by the galactic bar. We find a consistency between the star-forming region and the temperature inferred by the infrared color, suggesting that the distribution of subkiloparsec-scale temperature is driven by star formation.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Deep probabilistic direction prediction in 3D with applications to directional dark matter detectors

Abstract We present the first method to probabilistically predict 3D direction in a deep neural network model. The probabilistic predictions are modeled as a heteroscedastic von Mises-Fisher distribution on the sphere S 2 , giving a simple way to quantify aleatoric uncertainty. This approach generalizes the cosine distance loss which is a special case of our loss function when the uncertainty is assumed to be uniform across samples. We develop approximations required to make the likelihood function and gradient calculations stable. The method is applied to the task of predicting the 3D directions of electrons, the most complex signal in a class of experimental particle physics detectors designed to demonstrate the particle nature of dark matter and study solar neutrinos. Using simulated Monte Carlo data, the initial direction of recoiling electrons is inferred from their tortuous trajectories, as captured by the 3D detectors. For 40 keV electrons in a 70% He 30% CO 2 gas mixture at STP, the new approach achieves a mean cosine distance of 0.104 (26 ∘ ) compared to 0.556 (64 ∘ ) achieved by a non-machine learning algorithm. We show that the model is well-calibrated and accuracy can be increased further by removing samples with high predicted uncertainty. This advancement in probabilistic 3D directional learning could increase the sensitivity of directional dark matter detectors.

Computer Science↗
DOE OSTI2023

DSPS: Differentiable stellar population synthesis

ABSTRACT Models of stellar population synthesis (SPS) are the fundamental tool that relates the physical properties of a galaxy to its spectral energy distribution (SED). In this paper, we present DSPS: a python package for SPS. All of the functionality in DSPS is implemented natively in the JAX library for automatic differentiation, and so our predictions for galaxy photometry are fully differentiable, and directly inherit the performance benefits of JAX, including portability onto GPUs. DSPS also implements several novel features, such as i) a flexible empirical model for stellar metallicity that incorporates correlations with stellar age, ii) support for the Diffstar model that provides a physically-motivated connection between the star formation history of a galaxy (SFH) and the mass assembly of its underlying dark matter halo. We detail a set of theoretical techniques for using autodiff to calculate gradients of predictions for galaxy SEDs with respect to SPS parameters that control a range of physical effects, including SFH, stellar metallicity, nebular emission, and dust attenuation. When forward modelling the colours of a synthetic galaxy population, we find that DSPS can provide a factor of 5 speed-up over standard SPS codes on a CPU, and a factor of 300-400 on a modern GPU. When coupled with gradient-based techniques for optimization and inference, DSPS makes it practical to conduct expansive likelihood analyses of simulation-based models of the galaxy–halo connection that fully forward model galaxy spectra and photometry.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Differentiable Quantum Programming with Unbounded Loops

The emergence of variational quantum applications has led to the development of automatic differentiation techniques in quantum computing. Existing work has formulated differentiable quantum programming with bounded loops, providing a framework for scalable gradient calculation by quantum means for training quantum variational applications. However, promising parameterized quantum applications, e.g., quantum walk and unitary implementation, cannot be trained in the existing framework due to the natural involvement of unbounded loops. To fill in the gap, we provide the first differentiable quantum programming framework with unbounded loops, including a newly designed differentiation rule, code transformation, and their correctness proof. Technically, we introduce a randomized estimator for derivatives to deal with the infinite sum in the differentiation of unbounded loops, whose applicability in classical and probabilistic programming is also discussed. We implement our framework with Python and Q# and demonstrate a reasonable sample efficiency. Through extensive case studies, we showcase an exciting application of our framework in automatically identifying close-to-optimal parameters for several parameterized quantum applications.

Computer Science↗
DOE OSTI2022

Differentiable Predictions for Large Scale Structure with SHAMNet

In simulation-based models of the galaxy-halo connection, theoretical predictions for galaxy clustering and lensing are typically made based on Monte Carlo realizations of a mock universe. In this paper, we use Subhalo Abundance Matching (SHAM) as a toy model to introduce an alternative to stochastic predictions based on mock population, demonstrating how to make simulation-based predictions for clustering and lensing that are both exact and differentiable with respect to the parameters of the model. Conventional implementations of SHAM are based on iterative algorithms such as Richardson-Lucy deconvolution; here we use the JAX library for automatic differentiation to train SHAMNet, a neural network that accurately approximates the stellar-to-halo mass relation (SMHM) defined by abundance matching. In our approach to making differentiable predictions of large-scale structure, we map parameterized PDFs onto each simulated halo, and calculate gradients of summary statistics of the galaxy distribution by using autodiff to propagate the gradients of the SMHM through the statistical estimators used to measure one- and two-point functions. Our techniques are quite general, and we conclude with an overview of how they can be applied in tandem with more complex, higher-dimensional models, creating the capability to make differentiable predictions for the multi-wavelength universe of galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2016

A megacam survey of outer halo satellites. VI. The spatially resolved star-formation history of the carina dwarf spheroidal galaxy

We present the spatially resolved star-formation history (SFH) of the Carina dwarf spheroidal galaxy, obtained from deep, wide-field g and r imaging and a metallicity distribution from the literature. Our photometry covers ∼2 deg{sup 2}, reaching up to ∼10 times the half-light radius of Carina with a completeness higher than 50% at g ∼ 24.5, more than one magnitude fainter than the oldest turnoff. This is the first time a combination of depth and coverage of this quality has been used to derive the SFH of Carina, enabling us to trace its different populations with unprecedented accuracy. We find that Carina’s SFH consists of two episodes well separated by a star-formation temporal gap. These episodes occurred at old (>10 Gyr) and intermediate (2–8 Gyr) ages. Our measurements show that the old episode comprises the majority of the population, accounting for 54 ± 5% of the stellar mass within 1.3 times the King tidal radius, while the total stellar mass derived for Carina is 1.60±0.09×10{sup 6} M{sub ⊙}, and the stellar mass-to-light ratio is 1.8 ± 0.2. The SFH derived is consistent with no recent star formation, which hints that the observed blue plume is due to blue stragglers. We conclude that the SFH of Carina evolved independently of the tidal field of the Milky Way, since the frequency and duration of its star-formation events do not correlate with its orbital parameters. This result is supported by the age–metallicity relation observed in Carina and the gradients calculated indicating that outer regions are older and more metal-poor.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Numerical Analysis of Regular Material Point Method and its Application to Multiphase Flows

The material point method (MPM) is gaining wide popularity in engineering research to model and simulate complex multiphase flow dynamics. The method relies on solving the governing equations of motion and transport in a Lagrangian framework using particles also known as material points. The fluid and kinematic properties are stored on the material points while the spatial gradient calculation and temporal integration are performed on a background grid. This Lagrangian framework allows for large deformations, easy integration of constitutive models, and direct import of complex geometries as particles. However, despite their increasing popularity, very few studies have addressed the issues of numerical resolution and stability of MPM techniques. The presence of additional factors such as the number of material points-per-cell, the location of the material points, the CFL-like condition used in time update, and the grid shape functions also increase the complexity of the error analysis when compared to other finite element methods. In this presentation, we analyze the various forms of error incurred in the application of MPM to continuum mechanics and multiphase flows. The effect of the previously mentioned factors on the error dynamics is studied. The application of these principles to canonical and industrial problems is also presented.

high pressure reverse osmosis↗