SEARCH · Engineering Papers

Results for “Linear Regression”

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.

SourceResearch connectionFrom yearThrough year

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

1,233 records · Page 69

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Robust scalable initialization for Bayesian variational inference with multi-modal Laplace approximations

Predictive modeling typically relies on Bayesian model calibration to provide uncertainty quantification. Variational inference utilizing fully independent (“mean-field”) Gaussian distributions are often used as approximate probability density functions. This simplification is attractive since the number of variational parameters grows only linearly with the number of unknown model parameters. However, the resulting diagonal covariance structure and unimodal behavior can be too restrictive to provide useful approximations of intractable Bayesian posteriors that exhibit highly non-Gaussian behavior, including multimodality. High-fidelity surrogate posteriors for these problems can be obtained by considering the family of Gaussian mixtures. Gaussian mixtures are capable of capturing multiple modes and approximating any distribution to an arbitrary degree of accuracy, while maintaining some analytical tractability. Unfortunately, variational inference using Gaussian mixtures with full-covariance structures suffers from a quadratic growth in variational parameters with the number of model parameters. The existence of multiple local minima due to strong nonconvex trends in the loss functions often associated with variational inference present additional complications, These challenges motivate the need for robust initialization procedures to improve the performance and computational scalability of variational inference with mixture models. In this work, we propose a method for constructing an initial Gaussian mixture model approximation that can be used to warm-start the iterative solvers for variational inference. The procedure begins with a global optimization stage in model parameter space. In this step, local gradient-based optimization, globalized through multistart, is used to determine a set of local maxima, which we take to approximate the mixture component centers. Around each mode, a local Gaussian approximation is constructed via the Laplace approximation. Finally, the mixture weights are determined through constrained least squares regression. The robustness and scalability of the proposed methodology is demonstrated through application to an ensemble of synthetic tests using high-dimensional, multimodal probability density functions. Here, the practical aspects of the approach are demonstrated with inversion problems in structural dynamics.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1987

Right and left ventricular volumes in vitro by a new nongeometric method

We present an evaluation of a new nongeometric technique for calculating right and left ventricular volumes. This method calculates ventricular chamber volumes from multiple cross-sectional echocardiographic views taken from a single point as the echo beam is tilted progressively through the ventricle. Right and left ventricular volumes are calculated from both the approximate short axis and approximate apical position on 20 in vitro human hearts and compared with the actual chamber volumes. The results for both ventricles from both positions are excellent. Correlation coefficients are > 0.95 for all positions; the standard errors are in the range of 5 to 7 mL and the slopes and intercepts for the regression lines are not significantly different from 1 and 0, respectively (except for the left ventricular short-axis intercept). For all positions, approximately 6 to 8 views are needed for peak accuracy (7.5 degrees to 10 degrees separation). This approach offers several advantages. No geometric assumptions about ventricular shape are made. All images are acquired from a single point (or window), and the digitized points can be used to make a three-dimensional reconstruction of the ventricle. Also, during the calculations a volume distribution curve for the ventricle is produced. The shape of this curve can be characteristic for certain situations (ie, right ventricle, short axis) and can be used to make new simple equations for calculating volume. We conclude that this is an accurate nongeometric method for determining both right and left ventricular volumes in vitro.

NASA Discipline Cardiopulmonary↗

DOE OSTI2022

Machine learning changes the rules for flux limiters

Learning to integrate non-linear equations from highly resolved direct numerical simulations has seen recent interest for reducing the computational load for fluid simulations. Here, we focus on determining a flux-limiter for shock capturing methods. Focusing on flux limiters provides a specific plug-and-play component for existing numerical methods. Since their introduction, an array of flux limiters has been designed. Using the coarse-grained Burgers' equation, we show that flux-limiters may be rank-ordered in terms of their log-error relative to high-resolution data. We then develop a theory to find an optimal flux-limiter and present flux-limiters that outperform others tested for integrating Burgers' equation on lattices with [Formula: see text], and 2x, 3x, 4x, and 8x coarse-grainings. We train a continuous piecewise linear limiter by minimizing the mean-squared misfit to six-grid point segments of high-resolution data, averaged over all segments. While flux limiters are generally designed to have an output of φ(r) = 1 at a flux ratio of r = 1, our limiters are not bound by this rule and yet produce a smaller error than standard limiters. Here we find that our machine learned limiters have distinctive features that may provide new rules-of-thumb for the development of improved limiters. Additionally, we use our theory to learn flux-limiters that outperform standard limiters across a range of values (as opposed to at a specific fixed value) of coarse-graining, number of discretized bins, and diffusion parameter. This demonstrates the ability to produce flux limiters that should be more broadly useful than standard limiters for general applications.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS2001

Human interactions in space: results from Shuttle/Mir

Background: Anecdotal reports from space and results from simulation studies on Earth have suggested that space crewmembers may experience decrements in their interpersonal environment over time and may displace tension and dysphoria to mission control personnel. Methods: To evaluate these issues, we studied 5 American astronauts, 8 Russian cosmonauts, and 42 American and 16 Russian mission control personnel who participated in the Shuttle/Mir space program. Subjects completed questions from subscales of the Profile of Mood States, the Group Environment Scale, and the Work Environment Scale on a weekly basis before, during, and after the missions. Results: Among the crewmembers, there was little evidence for significant time effects based on triphasic (U-shaped) or linear models for the 21 subscales tested, although the presence of an initial novelty effect that declined over time was found in three subscales for the astronauts. Compared with work groups on Earth, the crewmembers reported less dysphoria and perceived their crew environment as more constraining, cohesive, and guided by leadership. There was no change in ratings of mood and interpersonal environment before, during, and after the missions. Conclusions: There was little support for the presence of a moderate to strong time effect that influenced the space crews. Crewmembers perceived their work environment differently from people on Earth, and they demonstrated equanimity in mood and group perceptions, both in space and on the ground. Grant numbers: NAS9-19411. c 2001. Elsevier Science Ltd. All rights reserved.

Soyuz Project↗

NASA NTRS2024

Parameterizing Spectral Surface Reflectance Relationships for the Dark Target Aerosol Algorithm Applied to A Geostationary Imager

Originally developed for the Moderate Resolution Imaging Spectroradiometer (MODIS) in polar, sun-synchronous low-earth orbit (LEO), the Dark Target (DT) aerosol retrieval algorithm relies on the assumption of a Surface Reflectance Parameterization (SRP) over land surfaces. Specifically for vegetated and dark-soiled surfaces, values of surface reflectance in blue and red visible-wavelength bands are assumed to be nearly linearly related to each other and to the value in a shortwave infrared (SWIR) wavelength band. This SRP also includes dependencies on scattering angle and a normalized difference vegetation index computed from two SWIR bands (NDVISWIR). As the DT retrieval algorithm is being ported to new sensors to continue and expand the aerosol data record, we assess whether the MODIS-assumed SRP can be used for these sensors. Here, we specifically assess SRP for the Advanced Baseline Imager (ABI) aboard, the Geostationary Operational Environmental Satellite (GOES)-16/East (ABIE). First, we find that using MODIS-based SRP leads to higher biases and artificial diurnal signatures in aerosol optical depth (AOD)retrievals from ABIE. The primary reason appears to be that geostationary orbit (GEO) encounters an entirely different set of observation geometry than does LEO, primarily with regards to solar angles coupled with fixed view angles. Therefore, we have developed a new SRP for GEO that draws the angular shape of the surface bidirectional reflectance. We also introduce modifications to the parametrization of both red-SWIR and blue-red spectral relationships to include additional information.The revised Red-SWIR SRP includes solar zenith angle, NDVISWIR, and land-type percentage from an ancillary database. The blue-red SRP adds dependencies on the scattering angle and NDVISWIR. The new SRPs improve the AOD retrieval of ABIE in terms of overall less bias and mitigation of the overestimation around local noon. The average bias of DT AOD compared to AERONET AOD shows a reduction from 0.08 to 0.03, while the bias of local solar noon decreases from 0.12 to 0.03.The agreement between DT and AERONET AOD is established through regression slope of 1.06 and y-intercept of 0.01with correlation coefficient is0.74. By using the new SRP, the percentage of data falling within the expected error range (±0.05 + 15%) is notably risen from 54% to 78%.

Mijin Kim↗

DOE OSTI2022

Molecular dipole moment learning via rotationally equivariant derivative kernels in molecular-orbital-based machine learning

This study extends the accurate and transferable molecular-orbital-based machine learning (MOB-ML) approach to modeling the contribution of electron correlation to dipole moments at the cost of Hartree–Fock computations. A MOB pairwise decomposition of the correlation part of the dipole moment is applied, and these pair dipole moments could be further regressed as a universal function of MOs. The dipole MOB features consist of the energy MOB features and their responses to electric fields. An interpretable and rotationally equivariant derivative kernel for Gaussian process regression (GPR) is introduced to learn the dipole moment more efficiently. The proposed problem setup, feature design, and ML algorithm are shown to provide highly accurate models for both dipole moments and energies on water and 14 small molecules. To demonstrate the ability of MOB-ML to function as generalized density-matrix functionals for molecular dipole moments and energies of organic molecules, we further apply the proposed MOB-ML approach to train and test the molecules from the QM9 dataset. The application of local scalable GPR with Gaussian mixture model unsupervised clustering GPR scales up MOB-ML to a large-data regime while retaining the prediction accuracy. In addition, compared with the literature results, MOB-ML provides the best test mean absolute errors of 4.21 mD and 0.045 kcal/mol for dipole moment and energy models, respectively, when training on 110 000 QM9 molecules. The excellent transferability of the resulting QM9 models is also illustrated by the accurate predictions for four different series of peptides.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2025

Accelerating phase field simulations through a hybrid adaptive Fourier neural operator with U-net backbone

Prolonged contact between a corrosive liquid and metal alloys can cause progressive dealloying. For one such process as liquid-metal dealloying (LMD), phase field models have been developed to understand the mechanisms leading to complex morphologies. However, the LMD governing equations in these models often involve coupled non-linear partial differential equations (PDE), which are challenging to solve numerically. In particular, numerical stiffness in the PDEs requires an extremely refined time step size (on the order of 10 -12 s or smaller). This computational bottleneck is especially problematic when running LMD simulation until a late time horizon is required. This motivates the development of surrogate models capable of leaping forward in time, by skipping several consecutive time steps at-once. In this paper, we propose a U-shaped adaptive Fourier neural operator (U-AFNO), a machine learning (ML) based model inspired by recent advances in neural operator learning. U-AFNO employs U-Nets for extracting and reconstructing local features within the physical fields, and passes the latent space through a vision transformer (ViT) implemented in the Fourier space (AFNO). We use U-AFNOs to learn the dynamics of mapping the field at a current time step into a later time step. We also identify global quantities of interest (QoI) describing the corrosion process (e.g., the deformation of the liquid-metal interface, lost metal, etc.) and show that our proposed U-AFNO model is able to accurately predict the field dynamics, in spite of the chaotic nature of LMD. Most notably, our model reproduces the key microstructure statistics and QoIs with a level of accuracy on par with the high-fidelity numerical solver, while achieving a significant 11, 200 × speed-up on a high-resolution grid when comparing the computational expense per time step. Finally, we also investigate the opportunity of using hybrid simulations, in which we alternate forward leaps in time using the U-AFNO with high-fidelity time stepping. We demonstrate that while advantageous for some surrogate model design choices, our proposed U-AFNO model in fully auto-regressive settings consistently outperforms hybrid schemes.

36 MATERIALS SCIENCE↗

NASA NTRS1997

Combined influences of gravitoinertial force level and visual field pitch on visually perceived eye level

Psychophysical measurements of the level at which observers set a small visual target so as to appear at eye level (VPEL) were made on 13 subjects in 1.0 g and 1.5 g environments in the Graybiel Laboratory rotating room while they viewed a pitched visual field or while in total darkness. The gravitoinertial force was parallel to the z-axis of the head and body during the measurements. The visual field consisted of two 58 degrees high, luminous, pitched-from-vertical, bilaterally symmetric, parallel lines, viewed in otherwise total darkness. The lines were horizontally separated by 53 degrees and presented at each of 7 angles of pitch ranging from 30 degrees with the top of the visual field turned away from the subject (top backward) to 30 degrees with the top turned toward the subject (top forward). At 1.5 g, VPEL changed linearly with the pitch of the 2-line stimulus and was depressed with top backward pitch and elevated with top forward pitch as had been reported previously at 1.0 g (1,2); however, the slopes of the VPEL-vs-pitch functions at 1.0 g and 1.5 g were indistinguishable. As reported previously also (3,4), the VPEL in darkness was considerably lower at 1.5 g than at 1.0 g; however, although the y-intercept of the VPEL-vs-pitch function in the presence of the 2-line visual field (visual field erect) was also lower at 1.5 g than at 1.0 g as it was in darkness, the G-related difference was significantly attenuated by the presence of the visual field. The quantitative characteristics of the results are consistent with a model in which VPEL is treated as a consequence of an algebraic weighted average or a vector sum of visual and nonvisual influences although the two combining rules lead to fits that are equally good.

Non-NASA Center↗

DOE OSTI2025

Nonlinearity of the post-spinel transition and its expression in slabs and plumes worldwide

Phase transitions in the mantle control its internal dynamics and structure. The post-spinel transition marks the upper–lower mantle boundary, where ringwoodite dissociates into bridgmanite plus ferropericlase, and its Clapeyron slope regulates mantle flow across it. This interaction has previously been assumed to have no lateral spatial variations, based on the assumption of a linear post-spinel boundary in pressure and temperature. Here we present laser-heated diamond anvil cell experiments with synchrotron X-ray diffraction to better constrain this boundary, especially at higher temperatures. Combining our data with results from the literature, and using a global analysis based on machine learning, we find a pronounced nonlinearity in the post-spinel boundary, with its slope ranging from –4 MPa/K at 2100 K, to –2 MPa/K at 1950 K, and to 0 MPa/K at 1600 K. Changes in temperature over time and space can therefore cause the post-spinel transition to have variable effects on mantle convection and the movement of subducting slabs and upwelling plumes.

58 GEOSCIENCES↗