Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Gaussian process model”

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 775 records · Page 43
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Bayesian Optimization of Catalysis with In-Context Learning

Large language models (LLMs) can perform accurate classification with zero or few examples through in-context learning (ICL), allowing the model to observe query-relevant examples at inference time and eliminating the need for additional weight updates to generalize beyond its original training data. We extend this capability to regression with uncertainty estimation using frozen LLMs (e.g., GPT-4o, Gemini), enabling Bayesian optimization (BO) in natural language without explicit model training or feature engineering. We apply this to materials discovery by representing materials as synthesis and testing procedures for use in natural language prompts. This Bayesian, design-first approach prioritizes optimization toward target material properties before detailed characterization, in contrast to conventional experimental workflows that often emphasize characterization of suboptimal materials. On benchmarks like aqueous solubility and oxidative coupling of methane (OCM), BO-ICL matches or outperforms Gaussian processes. In live experiments on the reverse water–gas shift (RWGS) reaction, BO-ICL identifies multimetallic catalysts that approach equilibrium CO yield within 6 and 10 iterations from a pool of 3,700 and 360,000 candidates, respectively. Our method redefines materials representation and accelerates discovery, with broad applications across catalysis, materials science, and AI.

Calibration↗
DOE OSTI2019

Design and Analysis of a Wake Steering Controller with Wind Direction Variability

Wind farm control strategies are being developed to mitigate wake losses in wind farms, increasing energy production. Wake steering is a type of wind farm control in which a wind turbine's yaw position is misaligned from the wind direction, causing its wake to deflect away from downstream turbines. Current modeling tools used to optimize and estimate energy gains from wake steering are designed to represent wakes for fixed wind directions. However, wake steering controllers must operate in dynamic wind conditions and a turbine's yaw position cannot perfectly track changing wind directions. Research has been conducted on robust wake steering control optimized for variable wind directions. In this paper, the design and analysis of a wake steering controller with wind direction variability is presented for a two-turbine array using the FLOw Redirection and Induction in Steady State (FLORIS) control-oriented wake model. First, the authors propose a method for modeling the turbulent and low-frequency components of the wind direction, where the slowly varying wind direction serves as the relevant input to the wake model. Next, we explain a procedure for finding optimal yaw offsets for dynamic wind conditions considering both wind direction and yaw position uncertainty. We then performed simulations with the optimal yaw offsets applied using a realistic yaw offset controller in conjunction with a baseline yaw controller, showing good agreement with the predicted energy gain using the probabilistic model. Using the Gaussian wake model in FLORIS as an example, we compared the performance of yaw offset controllers optimized for static and dynamic wind conditions for different turbine spacings and turbulence intensity values. For a spacing of 5 rotor diameters and a turbulence intensity of 10%, robust yaw offsets optimized for variable wind directions yielded an energy gain improvement of 128%. In general, accounting for wind direction variability in the yaw offset optimization process was found to improve energy production more as the separation distance increased, whereas the relative improvement remained roughly the same for the range of turbulence intensity values considered.

17 WIND ENERGY↗
NASA NTRS1996

The barberplaid illusion: plaid motion is biased by elongated apertures

The perceived direction of motion of plaids windowed by elongated spatial Gaussians is biased toward the window's long axis. The bias increases as the relative angle between the plaid motion and the long axis of the window increases, peaks at a relative angle of approximately 45 deg, and then decreases. The bias increases as the window is made narrower (at fixed height) and decreases as the component spatial frequency increases (at fixed aperture size). We examine several models of human motion processing (cross-correlation, motion-energy, intersection-of-constraints, and vector-sum), and show that none of these standard models can predict our data. We conclude that spatial integration of motion signals plays a crucial role in plaid motion perception and that current models must be explicitly expanded to include such spatial interactions.

NASA Center ARC↗
DOE OSTI2024

Probabilistic Analysis of Long-Term Degradation of Microwave Cavity Flow Sensor

We are investigating a microwave resonant cavity transducer for flow sensing in the vessel of a high temperature fluid advanced reactor (AR), such as a molten salt cooled reactor (MSCR) or a sodium fast reactor (SFR). This transducer is a hollow metallic cylindrical cavity, with the flat wall of the cylinder flexible enough to undergo microscopic deflection due to dynamic fluid pressure. Membrane deflection leads to a shift in the resonant frequency, which can be detected with a spectrum analyzer. We have performed a proof-of-concept experiment of flow sensing with the transducer in liquid sodium at 340°C in impinging liquid jet geometry. The transducer remained in liquid sodium for 70 days. After removal, no structural damage was observed, and the expected transducer response was verified in a water test. Because long-term (multi-year) experimental tests of transducer resilience to harsh environment are not practical, we have developed a probabilistic model of creep to estimate transducer resilience to the harsh environment. The probabilistic model considers diffusion creep under the condition of high temperature and low stress, where the stress and temperature are allowed to be random variables with Gaussian distributions. Using the probabilistic model, we estimate inelastic membrane deflections due to creep for several temperature ranges. We conclude that for temperatures less than 650°C, creep has negligible long-term effect on the transducer performance. Since a yellowish residue was observed on the transducer surface after 70 days of immersion in liquid sodium, we have investigated possible evidence of corrosion. Chromium depletion is a typical indicator of the corrosion process in stainless steel. Scraping off a residue from the transducer and performing scanning electron microscopy (SEM) with energy dispersive analysis (EDS) did not find any chromium in the residue. Approximately 60% of the residue consisted of copper, which can be attributed to contamination of sodium due to powder residue from machining of copper and brass components of the transducer.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
NASA NTRS2011

Filtering Methods for Error Reduction in Spacecraft Attitude Estimation Using Quaternion Star Trackers

Precision attitude determination for recent and planned space missions typically includes quaternion star trackers (ST) and a three-axis inertial reference unit (IRU). Sensor selection is based on estimates of knowledge accuracy attainable from a Kalman filter (KF), which provides the optimal solution for the case of linear dynamics with measurement and process errors characterized by random Gaussian noise with white spectrum. Non-Gaussian systematic errors in quaternion STs are often quite large and have an unpredictable time-varying nature, particularly when used in non-inertial pointing applications. Two filtering methods are proposed to reduce the attitude estimation error resulting from ST systematic errors, 1) extended Kalman filter (EKF) augmented with Markov states, 2) Unscented Kalman filter (UKF) with a periodic measurement model. Realistic assessments of the attitude estimation performance gains are demonstrated with both simulation and flight telemetry data from the Lunar Reconnaissance Orbiter.

Calhoun, Philip C.↗
NASA NTRS1971

The role and use of the stochastic linear-quadratic-Gaussian problem in control system design.

The role of the linear-quadratic stochastic control problem in engineering design is reviewed in tutorial fashion. The design approach is motivated by considering the control of a nonlinear uncertain plant about a desired input-output response. It is demonstrated how a design philosophy based on (1) deterministic perturbation control, (2) stochastic state estimation, and (3) linearized stochastic control leads to an overall closed-loop control system. The emphasis of the paper is on the philosophy of the design process, the modeling issue, and the formulation of the problem; the results are given for the sake of completeness, but no proofs are included. The systematic off-line nature of the design process is stressed throughout.

Athans, M.↗
DOE OSTI2025

Compactly‐Supported Nonstationary Kernels for Computing Exact Gaussian Processes on Big Data

The Gaussian process (GP) is a widely used method for analyzing large-scale data sets, including spatio-temporal measurements of nonlinear processes that are now commonplace in the environmental sciences. Traditional implementations of GPs involve stationary kernels (also termed covariance functions) that limit their flexibility, and exact methods for inference that prevent application to data sets with more than about 10,000 points. Modern approaches to address stationarity assumptions generally fail to accommodate large data sets, while all attempts to address scalability focus on approximating the Gaussian likelihood, which can involve subjectivity and lead to inaccuracies. In this work, we explicitly derive an alternative kernel that can discover and encode both sparsity and nonstationarity. We embed the kernel within a fully Bayesian GP model and leverage high-performance computing resources to enable the analysis of massive data sets. We demonstrate the favorable performance of our novel kernel relative to existing exact and approximate GP methods across a variety of synthetic data examples. Furthermore, we conduct space–time prediction based on more than 1 million measurements of daily maximum temperature and verify that our results outperform state-of-the-art methods in the Earth sciences. More broadly, having access to exact GPs that use ultra-scalable, sparsity-discovering, nonstationary kernels allows GP methods to truly compete with a wide variety of machine learning methods.

Gaussian processes↗
NASA NTRS2021

The Asymmetric Inner Disk of the Herbig Ae Star HD 163296 in the Eyes of VLTI/MATISSE: Evidence for a Vortex?

Context.A complex environment exists in the inner few astronomical units of planet-forming disks. High-angular-resolution observa-tions play a key role in our understanding of the disk structure and the dynamical processes at work.Aims.In this study we aim to characterize the mid-infrared brightness distribution of the inner disk of the young intermediate-massstar HD 163296 from early VLTI/MATISSE observations taken in theL- andN-bands. We put special emphasis on the detection ofpotential disk asymmetries.Methods.We use simple geometric models to fit the interferometric visibilities and closure phases. Our models include a smoothedring, a flat disk with an inner cavity, and a 2D Gaussian. The models can account for disk inclination and for azimuthal asymmetriesas well. We also perform numerical hydrodynamical simulations of the inner edge of the disk.Results.Our modeling reveals a significant brightness asymmetry in theL-band disk emission. The brightness maximum of the asym-metry is located at the NW part of the disk image, nearly at the position angle of the semimajor axis. The surface brightness ratio inthe azimuthal variation is3.5±0.2. Comparing our result on the location of the asymmetry with other interferometric measurements,we confirm that the morphology of ther<0.3au disk region is time-variable. We propose that this asymmetric structure, located in ornear the inner rim of the dusty disk, orbits the star. To find the physical origin of the asymmetry, we tested a hypothesis where a vortexis created by Rossby wave instability, and we find that a unique large-scale vortex may be compatible with our data. The half-lightradius of theL-band-emitting region is0.33±0.01au, the inclination is52◦+5◦−7◦, and the position angle is143◦±3◦. Our models predictthat a non-negligible fraction of theL-band disk emission originates inside the dust sublimation radius forμm-sized grains. Refractorygrains or large (&10μm-sized) grains could be the origin of this emission.N-band observations may also support a lack of smallsilicate grains in the innermost disk (r.0.6au), in agreement with our findings fromL-band data.

J Varga↗
DOE OSTI2023

Adaptive Deep Reinforcement Learning Algorithm for Distribution System Cyber Attack Defense With High Penetration of DERs

With grid modernization, smart inverters are increasingly used to execute advanced controls for distribution network reliability. However, this also increases the cyber-attack space. Here this paper focuses on the defense approaches to restore the system to normal operation circumstances in the presence of cyber-attacks. A unique deep reinforcement learning (DRL) method is developed to minimize voltage violations and reduce power losses for impacted feeders. The defense problem is reformulated as a Markov decision-making process to dynamically control DERs while minimizing load shedding. This is achieved via an improved soft actor-critic (SAC)-based DRL algorithm, which can govern DER set points and load-shedding scenarios in discrete and continuous modes via the auto-tune entropy and Gaussian policy features. Numerical comparison results on the modified IEEE 123-node system with other control approaches, such as Volt-VAR (VV), Volt-Watt (VW), and model predictive control (MPC) show that the proposed method can eliminate voltage violations and provide feasible control actions that perform complete mitigation of cyber-threats.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

Gaussian processes for inferring parton distributions

The extraction of parton distribution functions (PDFs) from experimental or lattice QCD data is an ill-posed inverse problem, where regularization strongly impacts both systematic uncertainties and the reliability of the results. We study a framework based on Gaussian Process Regression (GPR) to reconstruct PDFs from lattice QCD matrix elements. Within a Bayesian framework, Gaussian processes serve as flexible priors that encode uncertainties, correlations, and constraints without imposing rigid functional forms. We investigate a wide range of kernel choices, mean functions, and hyperparameter treatments. We quantify information gained from the data using the Kullback-Leibler divergence. Synthetic data tests demonstrate the consistency and robustness of the method. Our study establishes GPR as a systematic and non-parametric approach to PDF reconstruction, offering controlled uncertainty estimates and reduced model bias in lattice QCD analyses.

hadronic spectroscopy↗
DOE OSTI2022

Improved $^{95}\mathrm{Mo}$ neutron resonance parameters and astrophysical reaction rates

We report improved 95 Mo neutron resonance parameters and reaction rates are important for nuclear astrophysics, testing nuclear models, and nuclear criticality safety. However, despite many previous neutron-capture and total cross-section measurements on this nuclide, there still is much room for improvement as well as several discrepancies. For example, there are very few firm resonance spin and parity assignments; average resonance parameters are available only for each parity, the currently recommended astrophysical reaction rate results in disagreements between stellar models and meteoric isotopic anomalies, and there are substantial disagreements in the neutron-capture cross section at low energies important for nuclear criticality safety. To obtain an improved set of neutron resonance parameters and astrophysical reaction rates for 95 Mo. High-resolution neutron-capture and transmission data were measured at the Oak Ridge Electron Linear Accelerator (ORELA) using highly isotopically enriched 95 Mo samples. The neutron-capture apparatus, data reduction, and analysis were improved so that information contained in the γ-ray cascade following neutron capture were used to assign resonance J π values. Following this, simultaneous analysis of the new neutron-capture and transmission data was used to obtain resonance energies, gamma widths, and neutron widths and their uncertainties to a maximum energy of 10 keV. Accurate neutron-capture cross sections also were obtained for the unresolved resonance region to a maximum energy of 500 keV and, together with the new resonance parameters, used to calculate the astrophysical reaction rates in the temperature range from 5 to 30 keV. A vastly improved set of 95Mo neutron resonance parameters and an astrophysical reaction rate accurate to about 3% were obtained. Firm J π assignments were determined for 261 of the 314 observed resonances. This is a very large improvement over the previously published 32 firm J π assignments for 108 resonances. Also, the number of resonances having both firm J π assignments and Γ γ values was increased by almost a factor of 24—from 11 to 261. Neutron- and total-radiation-width distributions and average resonance spacings, average total radiation widths, and neutron strength functions were obtained for the six different s- and p-wave possibilities. Parameters for the lowest s-wave resonance, which is most important for criticality benchmarks, were obtained with high accuracy. Simple modification of the neutron-capture apparatus and expansion and improvement of data analysis techniques led to a large increase in firm J π assignments for 95 Mo neutron resonances. The resulting astrophysical reaction rate is 20%–30% larger than the currently recommended rate at s-process temperatures, which should lead to better agreement between stellar models and meteoric isotopic anomalies. The neutron-capture cross section at low energies is substantially larger than recommended in the latest evaluation, which is problematical for criticality benchmarks. The average resonance spacing as a function of spin and parity is significantly different from current models. The total-radiation-width distributions are significantly broader than predicted by theory and show significant departures from the expected Gaussian shapes.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS1990

Incorporation of shuttle CCT parameters in computer simulation models

Computer simulations of shuttle missions have become increasingly important during recent years. The complexity of mission planning for satellite launch and repair operations which usually involve EVA has led to the need for accurate visibility and access studies. The PLAID modeling package used in the Man-Systems Division at Johnson currently has the necessary capabilities for such studies. In addition, the modeling package is used for spatial location and orientation of shuttle components for film overlay studies such as the current investigation of the hydrogen leaks found in the shuttle flight. However, there are a number of differences between the simulation studies and actual mission viewing. These include image blur caused by the finite resolution of the CCT monitors in the shuttle and signal noise from the video tubes of the cameras. During the course of this investigation the shuttle CCT camera and monitor parameters are incorporated into the existing PLAID framework. These parameters are specific for certain camera/lens combinations and the SNR characteristics of these combinations are included in the noise models. The monitor resolution is incorporated using a Gaussian spread function such as that found in the screen phosphors in the shuttle monitors. Another difference between the traditional PLAID generated images and actual mission viewing lies in the lack of shadows and reflections of light from surfaces. Ray tracing of the scene explicitly includes the lighting and material characteristics of surfaces. The results of some preliminary studies using ray tracing techniques for the image generation process combined with the camera and monitor effects are also reported.

Huntsberger, Terry↗
DOE OSTI2023

Improving resolution and resolvability of single-particle cryoEM structures using Gaussian mixture models

Cryogenic electron microscopy is widely used in structural biology, but its resolution is often limited by the dynamics of the macromolecule. Here, we developed a refinement protocol based on Gaussian mixture models that integrates particle orientation and conformation estimation, and improves the alignment for flexible domains of protein structures. We demonstrated this protocol on multiple datasets, resulting in improved resolution and resolvability, locally and globally, by visual and quantitative measures.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Assessments of epistemic uncertainty using Gaussian stochastic weight averaging for fluid-flow regression

Here, we use Gaussian stochastic weight averaging (SWAG) to assess the epistemic uncertainty associated with neural-network-based function approximation relevant to fluid flows. SWAG approximates a posterior Gaussian distribution of each weight, given training data, and a constant learning rate. Having access to this distribution, it is able to create multiple models with various combinations of sampled weights, which can be used to obtain ensemble predictions. The average of such an ensemble can be regarded as the 'mean estimation', whereas its standard deviation can be used to construct 'confidence intervals', which enable us to perform uncertainty quantification (UQ) with regard to the training process of neural networks. We utilize representative neural-network-based function approximation tasks for the following cases: (i) a two-dimensional circular-cylinder wake; (ii) the DayMET dataset (maximum daily temperature in North America); (iii) a three-dimensional square-cylinder wake; and (iv) urban flow, to assess the generalizability of the present idea for a wide range of complex datasets. SWAG-based UQ can be applied regardless of the network architecture, and therefore, we demonstrate the applicability of the method for two types of neural networks: (i) global field reconstruction from sparse sensors by combining convolutional neural network (CNN) and multi-layer perceptron (MLP); and (ii) far-field state estimation from sectional data with two-dimensional CNN. We find that SWAG can obtain physically-interpretable confidence-interval estimates from the perspective of epistemic uncertainty. This capability supports its use for a wide range of problems in science and engineering.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

CFD-DEM solution verification: Fixed-bed studies

Fixed (frozen) particle configurations are used to study CFD-DEM convergence behavior in the limit of infinite grid resolution. A filtering scheme based on solving a diffusion equation, i.e., a Gaussian kernel, is used to interpolate particle information to the continuum grid, free of cell size limitations. A regression method is used for Richardson extrapolation of the discrete solutions to the grid-free solution. The regression method is shown to predict accurate extrapolation curves in the presence of the observed oscillatory convergence behavior. The extrapolation curve is used to complete the solution verification process by quantifying the numerical uncertainty present in the calculations. Several numerical and model parameters are studied including the spatial discretization method, the kernel width and the mean drag law. Here, a manageable level of discretization error is found with relative discretization errors generally <4%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Energy cost of localization of relational quantum information

Entanglement of spatially separated quantum states is usually defined with respect to a reference frame provided by some external observer. Thus, if one wishes to localize the quantum information within a spatially separated entangled state, one must enact an entanglement extraction protocol also defined with respect to that external frame. Entanglement extraction for Gaussian ground states in such an external frame construction has been shown to require a minimum energy and is hence an interesting process for gravitational physics, where examinations of localization vs energy cost have a long history. General covariance, however, precludes dependence on external frames. In order to enact an extraction protocol in a generally covariant theory, dependence on the external reference frame must first be removed and the states made relational. Here, we examine the implementation of an extraction protocol for Gaussian states, whose center of mass and relational degrees of freedom are entangled, in a relational toy model where translation invariance stands in for full diffeomorphism invariance. Constructing fully relational states and the corresponding extraction/localization can, in principle, be done in two ways. External frame position information can be removed through 𝐺-twirling over translations or one can spontaneously break the translation symmetry via the gradient of an auxiliary field, or 𝑍-model. We determine the energetics of quantum information localization after the states have been made fully relational via both the 𝐺-twirl and 𝑍-model. We also show one can obtain the 𝐺-twirl construction from a 𝑍-model as a limit of positive operator valued measurements.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Multifidelity Ensemble Kalman Filtering Using Surrogate Models Defined by Theory-Guided Autoencoders

Data assimilation is a Bayesian inference process that obtains an enhanced understanding of a physical system of interest by fusing information from an inexact physics-based model, and from noisy sparse observations of reality. The multifidelity ensemble Kalman filter (MFEnKF) recently developed by the authors combines a full-order physical model and a hierarchy of reduced order surrogate models in order to increase the computational efficiency of data assimilation. The standard MFEnKF uses linear couplings between models, and is statistically optimal in case of Gaussian probability densities. This work extends the MFEnKF into to make use of a broader class of surrogate model such as those based on machine learning methods such as autoencoders non-linear couplings in between the model hierarchies. We identify the right-invertibility property for autoencoders as being a key predictor of success in the forecasting power of autoencoder-based reduced order models. We propose a methodology that allows us to construct reduced order surrogate models that are more accurate than the ones obtained via conventional linear methods. Numerical experiments with the canonical Lorenz'96 model illustrate that nonlinear surrogates perform better than linear projection-based ones in the context of multifidelity ensemble Kalman filtering. We additionality show a large-scale proof-of-concept result with the quasi-geostrophic equations, showing the competitiveness of the method with a traditional reduced order model-based MFEnKF.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Reliability estimation of an advanced nuclear fuel using coupled active learning, multifidelity modeling, and subset simulation

Tristructural isotropic (TRISO)-coated particle fuel is a robust nuclear fuel and determining its reliability is critical for the success of advanced nuclear technologies. However, TRISO failure probabilities are small and the associated computational models are expensive. Here we used coupled active learning, multifidelity modeling, and subset simulation to estimate the failure probabilities of TRISO fuels using several 1D and 2D models. With multifidelity modeling, we replaced expensive high-fidelity (HF) model evaluations with information fusion from two low-fidelity (LF) models. For the 1D TRISO models, we considered three multifidelity modeling strategies: only Kriging, Kriging LF prediction plus Kriging correction, and deep neural network (DNN) LF prediction plus Kriging correction. While the results across these multifidelity modeling strategies compared satisfactorily, strategies employing information fusion from two LF models called the HF model least often. Next, for the 2D TRISO model, we considered two multifidelity modeling strategies: DNN LF prediction plus Kriging correction (data-driven) and 1D TRISO LF prediction plus Kriging correction (physics-based). The physics-based strategy, as expected, consistently required the fewest calls to the HF model. However, the data-driven strategy had a lower overall simulation time since the DNN predictions are instantaneous, and the 1D TRISO model requires a non-negligible simulation time.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗