Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “gaussian processes”

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 721 records · Page 40
All sourcesNASA NTRSDOE OSTI
NASA NTRS2015

Predicting Time Series Outputs and Time-to-Failure for an Aircraft Controller Using Bayesian Modeling

Safety of unmanned aerial systems (UAS) is paramount, but the large number of dynamically changing controller parameters makes it hard to determine if the system is currently stable, and the time before loss of control if not. We propose a hierarchical statistical model using Treed Gaussian Processes to predict (i) whether a flight will be stable (success) or become unstable (failure), (ii) the time-to-failure if unstable, and (iii) time series outputs for flight variables. We first classify the current flight input into success or failure types, and then use separate models for each class to predict the time-to-failure and time series outputs. As different inputs may cause failures at different times, we have to model variable length output curves. We use a basis representation for curves and learn the mappings from input to basis coefficients. We demonstrate the effectiveness of our prediction methods on a NASA neuro-adaptive flight control system.

Statistics↗
NASA NTRS2016

A Comparison of Metamodeling Techniques via Numerical Experiments

This paper presents a comparative analysis of a few metamodeling techniques using numerical experiments for the single input-single output case. These experiments enable comparing the models' predictions with the phenomenon they are aiming to describe as more data is made available. These techniques include (i) prediction intervals associated with a least squares parameter estimate, (ii) Bayesian credible intervals, (iii) Gaussian process models, and (iv) interval predictor models. Aspects being compared are computational complexity, accuracy (i.e., the degree to which the resulting prediction conforms to the actual Data Generating Mechanism), reliability (i.e., the probability that new observations will fall inside the predicted interval), sensitivity to outliers, extrapolation properties, ease of use, and asymptotic behavior. The numerical experiments describe typical application scenarios that challenge the underlying assumptions supporting most metamodeling techniques.

Crespo, Luis G.↗
NASA NTRS2018

Space-Filling Designs for Multi-Layer Nested Factors

This articles considers computer experiments where levels for continuous factors are selected in sequential order with the level selected for one factor directly a ecting the range of possible levels for the nested factor, and so on for a nite number of factors. In addition, we assume the nested relationships between the factors have no closed form solution. In this paper, we propose an approach for constructing a multi-layer nested factor design, or multi-NFD for short. This space- lling design approach takes advan- tage of the maximin criterion and can be analyzed using a standard Gaussian process model. While the multi-NFD approach can be adapted for future computer experi- ments involving factor relationships of this type, we present results from a particular aerospace computer simulation study.

Gryder, Ryan W.↗
NASA NTRS2019

A ~60 day Super-orbital Period Originating from the Ultraluminous X-ray Pulsar in M82

Ultraluminous X-ray (ULX) pulsars are a new class of object powered by apparent super-critical accretion onto magnetized neutron stars. Three sources in this class have been identified so far; M82X-2, NGC 5907 ULX-1, and NGC 7793 P13 have been found to have two properties in common; ∼1 s spin periods, and for NGC 5907 ULX-1 and NGC 7793 P13 periodic X-ray flux modulations on timescales of ∼60–80 days. M82 X-2 resides in a crowded field that includes the ULX M82 X-1 separated from X-2 by 5″, as well as other bright point sources. A 60 day modulation has been observed from the region, but the origin has been difficult to identify; both M82 X-1 and X-2 have been suggested as the source. In this paper we present the analysis of a systematic monitoring campaign by Chandra, the only X-ray telescope capable of resolving the crowded field. From a simple Lomb–Scargle periodogram analysis and a more sophisticated Gaussian Process analysis we find that only X-2 exhibits a periodic signal around 60 days, supporting previous claims that it is the origin. We also construct a phase-averaged flux profile of the modulations from higher-cadence Swift/XRT data and find that the flux variations in the Chandra data are fully consistent with the flux profile. Since the orbit of the neutron star and its companion is known to be 2.5 days, the ∼60 day period must be super-orbital in origin. The flux of the modulations varies by a factor of ∼100 from the minimum to the maximum, with no evidence for spectral variations, making the origin difficult to explain.

Individual galaxies↗
NASA NTRSDate not supplied

Uncertainty Propagation in Pre-Flight Prediction of Unmanned Aerial Vehicle Separation Violation

Current forecasts on the future of aeronautics suggest an in- creasing number of unmanned aerial vehicles entering the low- altitude airspace in the next decades (FAA, 2018; Kopardekar et al., 2016). Small vehicles for package delivery as well as larger vehicles for urban air mobility will change the airspace drastically, increasing density of operations both in time, i.e. high number of take-off and landings per unit time, and in space, operating in dense urban environment. This scenario poses challenges to the current approach to air traffic control, and large efforts from academia, industry and regulatory bodies are dedicated to the development of new traffic management strategies that leverage higher computing and simulating capabilities available today. In this paper, we propose a simple look-ahead approach to predict potential minimum separation violations at the strategic level, that is before vehicles start flying, depending on the predefined 4D trajectories and uncertainty affecting the wind acting along those routes. The wind field is extracted from the NOAA North America Mesoscale Forecast System and interpolated using Gaussian process regression, while uncertainty affecting the expected cruise airspeed is propagated through error intervals. The approach allows the prediction of aircraft separation as a function of time, highlighting potential safety violations that would go undetected if uncertainty affecting the expected 4D trajectories is not considered. The paper will also discuss issues related to accuracy and scalability of the approach to multiple vehicle operations.

Trajectory Prediction↗
NASA NTRS2021

The M-dwarf Ultraviolet Spectroscopic Sample. I. Determining Stellar Parameters for Field Stars

Accurate stellar properties are essential for precise stellar astrophysics and exoplanetary science. In the M-dwarf regime, much effort has gone into defining empirical relations that can use readily accessible observables to assess physical stellar properties. Often, these relations for the quantity of interest are cast as a nonlinear function of available data; in Bayesian modeling, however, the reverse is needed. In this article, we introduce a new Bayesian framework to self-consistently and simultaneously apply multiple empirical calibrations to fully characterize the mass, luminosity, radius, and effective temperature of a field age M-dwarf. This framework includes a new M-dwarf mass–radius relation with a scatter of 3.1% at fixed mass. We further introduce the M-dwarf Ultraviolet Spectroscopic Sample (MUSS), and apply our methodology to provide consistent stellar parameters for these nearby low-mass stars, selected as having available spectroscopic data in the ultraviolet. These targets are of interest largely as either exoplanet hosts or benchmarks in multiwavelength stellar activity. We use the field MUSS stars to define a low-mass main sequence in the solar neighborhood through Gaussian Process (GP) regression. These results enable us to empirically measure a feature in the GP derivative at M ⊙ that indicates where the MUSS transitions from fully to partly convective interiors.

J. Sebastian Pineda↗
NASA NTRSDate not supplied

Probability of Obstacle Collision for UAVs in Presence of Wind

For incorporation of unmanned aerial vehicles into the National Airspace, ensuring safety of the airspace including the vehicles, people, and property on the ground is of utmost importance. One of the safety-critical factors for unmanned aviation flights is the risk of deviating from a planned trajectory resulting in a variety of hazards, including potential loss of separation between vehicle and obstacles or unexpected battery energy consumption. Off-nominal conditions introduced by component failures, degraded controllability and environmental disturbances such as wind gusts can lead to unacceptable unexpected deviations from the flight trajectory. It is essential to accurately model such effects on the flight trajectory while computing safety thresholds such as minimum separation from surrounding obstacles, available battery resource to complete the mission or determining delay in the expected time of arrival of flights. In this paper, a tool is presented based on Gaussian Process Regression for wind representation over a pre-defined trajectory for fast, yet approximated, in-time evaluation of possible trajectory deviations caused by wind gusts. The deviation in the planned trajectory caused by wind is further simulated utilizing a 6 degrees-of-freedom (DOF) UAV trajectory simulator comprising of a rotorcraft lumped-mass model with LQRI controller. Both steady-state wind and wind gust effects are investigated. The probability of collision with obstacle is computed and demonstrated on real flight data from experimental flights of an octocopter at NASA Langley Research Center in the presence of simulated obstacles and wind conditions. Effect of varying wind conditions and varying UAV airspeed is further demonstrated on experimental flights in the presence of wind measured by ground based weather service stations. The proposed approach would eventually benefit timely mitigation of current and future safety-critical events in autonomous systems by enabling risk-informed decision making.

Portia Banerjee↗
NASA NTRS2016

EVEREST: Pixel Level Decorrelation of K2 Light Curves

We present EPIC Variability Extraction and Removal for Exoplanet Science Targets (EVEREST), an open-source pipeline for removing instrumental noise from K2 light curves. EVEREST employs a variant of pixel level decorrelation to remove systematics introduced by the spacecraft’s pointing error and a Gaussian process to capture astrophysical variability. We apply EVEREST to all K2 targets in campaigns 0–7, yielding light curves with precision comparable to that of the original Kepler mission for stars brighter than K(sub p) ≈ 13, and within a factor of two of the Kepler precision for fainter targets. We perform cross-validation and transit injection and recovery tests to validate the pipeline and compare our light curves to the other de-trended light curves available for download at the MAST High Level Science Products archive. We find that EVEREST achieves the highest average precision of any of these pipelines for unsaturated K2 stars. The improved precision of these light curves will aid in exoplanet detection and characterization, investigations of stellar variability, asteroseismology, and other photometric studies. The EVEREST pipeline can also easily be applied to future surveys, such as the TESS mission, to correct for instrumental systematics and enable the detection of low signal-to-noise transiting exoplanets. The EVEREST light curves and the source code used to generate them are freely available online.

Catalogs↗
NASA NTRSDate not supplied

PM 2.5 Concentrations over Major Metropolitan Regions Inferred from Airborne High Spectral Resolution Lidar Measurements Using Machine Learning Regression

We use measurements of near-surface aerosol backscatter, extinction, and depolarization acquired by four NASA Langley Research Center airborne High Spectral Resolution Lidars (HSRLs) to develop a machine learning regression methodology to infer PM2.5 concentrations at the surface and aloft. These airborne HSRL measurements were acquired over major metropolitan regions in the United States and Asia during more than 170 flights since 2010. Hourly surface PM2.5 measurements from the EPA air quality system and similar networks in other countries acquired within 10 km and 15 minutes of these near-surface HSRL measurements are used to train models that compute PM2.5 concentrations from the HSRL measurements. We examine several regression methods and find that exponential Gaussian Process algorithms consistently give the best performance in terms of the lowest root-mean-square (RMS) errors and the highest correlations. Model performance varies significantly depending on various combinations of HSRL aerosol measurements (e.g., aerosol backscatter, extinction, depolarization, backscatter color ratios, lidar ratios, aerosol optical thickness) and retrievals (e.g., mixed layer height, aerosol type) used in the regressions. Models that use near-surface measurements of aerosol backscatter and aerosol intensive properties such as depolarization, backscatter color ratio, and lidar ratio typically give the best performance with RMS errors around 4 mg/m3 and correlation coefficients above 0.9. HSRL measurements were often acquired when the aircraft flew systematic “raster-scan” patterns for several hours over these cities. These flight patterns enabled measurements of the spatial, temporal, and vertical variabilities in the distributions of aerosol backscatter and aerosol intensive properties and allowed us to derive the corresponding variabilities in PM2.5 concentrations. We present examples of such variabilities over urban areas in the United States as well as Asia. We describe also how the distribution of surface PM2.5 varies with aerosol type and use these retrievals to examine model simulations of surface PM2.5 in these metropolitan regions. We also discuss how this methodology may be applied to measurements from satellite lidars such as CALIOP on CALIPSO and ATLID on EarthCARE.

lidar↗
DOE OSTI2021

Towards Efficient Uncertainty estimation in deep learning for robust energy prediction in crystal materials

In material science, recent studies have started to explore the potential of using deep learning to improve property prediction from high-fidelity simulations, e.g, density functional theory (DFT). However, the design spaces are sometimes too large and intractable to sample completely. This results in a critical question that is how to evaluate the confidence and robustness of the prediction. In this paper, we propose an efficient approach to estimate uncertainty in deep learning using a single forward pass and then apply it for robust prediction of the total energy in crystal lattice structures. Our approach is built upon the deep kernel learning (DKL) that originally introduces to leverage the expressiveness of deep neural networks as input with a probabilistic prediction of Gaussian processes (GPs) as output. Existing DKL methods have difficulties in the accuracy of predictive uncertainty, training stability, and scaling to large datasets, which lead to significant barriers in real-world applications. We propose to address these challenges by using an inducing point approximate GP in feature space combined with spectral normalization as a regularization. We finally demonstrate our robust performance on an artificial example and a real-world application from materials chemistry.

Bi, Sirui↗
DOE OSTI2021

Testing Surrogate-Based Optimization with the Fortified Branin-Hoo Extended to Four Dimensions

Some popular functions used to test global optimization algorithms have multiple local optima, all with the same value, making them all global optima. It is easy to make them more challenging by fortifying them via adding a localized bump at the location of one of the optima. In previous work the authors illustrated this for the Branin-Hoo function and the popular differential evolution algorithm, showing that the fortified Branin-Hoo required an order of magnitude more function evaluations. This paper examines the effect of fortifying the Branin-Hoo function on surrogate- based optimization, which usually proceeds by adaptive sampling. Two algorithms are considered. The EGO algorithm, which is based on a Gaussian process (GP) and an algorithm based on radial basis functions (RBF). EGO is found to be more frugal in terms of the number of required function evaluations required to identify the correct basin, but it is expensive to run on a desktop, limiting the number of times the runs could be repeated to establish sound statistics on the number of required function evaluations. The RBF algorithm was cheaper to run, providing more sound statistics on performance. A four-dimensional version of the Branin-Hoo function was introduced in order to assess the effect of dimensionality. Furthermore, it was found that the difference between the ordinary function and the fortified one was much more pronounced for the four-dimensional function compared to the two dimensional one.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Leverage Score Sampling for Parametric PDEs (Final Technical Report)

This final technical report summarizes the accomplishments of work performed under DOE Office of Science Award DE-SC0022266, which is titled “Leverage Score Sampling for Parametric PDEs”. The goal of the project was to extend methods from Randomized Numerical Linear Algebra (RandNLA) to tackle central computational challenges in model order reduction and uncertainty quantification (UQ) for parametric partial differential equations (PDEs). In particular, we sought to use importance sampling methods originally developed for RandNLA to develop sample efficient active learning algorithms for approximating high-dimensional scalar functions, e.g. by polynomials, Gaussian process models, and simple neural networks. Such methods can be immediately applied to developing surrogate models or to approximating quantity of interest (QoI) surfaces. In the context of PDEs, each sample used for learning equates to the solution of the differential equation for a particular set of parameters, so sample efficiency translates to improved computational efficiency for a variety of downstream tasks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Adaptive, Active Learning, and Multifidelity Monte Carlo Methods in the MOOSE Stochastic Tools Module

MOOSE is an open-source computational platform for constructing multi-physics models and executing them in a massively parallel fashion. It has a stochastic tools module (STM) for forward/inverse uncertainty quantification (UQ) and surrogate modeling. This presentation details some recent developments to the STM with respect to the implementation of adaptive, active learning, and multifidelity Monte Carlo methods for forward UQ of computational models. Specifically, the adaptive Monte Carlo methods include Markov Chain Monte Carlo (MCMC)-driven algorithms like adaptive importance sampling and parallelized subset simulation for statistical QoI estimation, rare events analysis, and stochastic gradient-free optimization. The active learning methods include Gaussian Process (GP) surrogates and their training via Adam optimization, design of acquisition functions, and integration with samplers like Monte Carlo, adaptive importance, and parallelized subset simulation. These active learning methods are also designed to work in a batch mode, wherein, the required calls to the full computational model are executed in parallel whenever a user-specified batch size is met. The multifidelity methods in STM are broadly divided into two categories: hierarchical, where a defined hierarchy exists among the low-fidelity models, and peer, where all the low-fidelity models are treated equally. A GP surrogate is used to learn the differences between the low- and high-fidelity models in both multifidelity categories, and acquisition functions from the active learning classes are used to decide whether to rely on a low-fidelity model or call the expensive high-fidelity model. Alongside the software description and usage, applications are also presented to nuclear engineering computational models including a TRISO nuclear fuel particle, a reactor pressure vessel, and a heat-pipe microreactor.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Accelerating Thermochemical Equilibrium Calculations for Nuclear Reactor Applications

Thermochemical properties play a key role in modeling and simulation of several key phenomena in nuclear reactors. There has been an increasing interest in incorporating CALPHAD-based formulations in multiphysics simulations including for Molten Salt Reactors where knowledge of phase evolution of the salt and the chemical potentials of various elements are of utmost importance in source term analyses and redox control. However, the size of such simulations is often limited by the high computational cost of full thermodynamic equilibrium calculations. This work discusses the current efforts aimed at accelerating thermochemical equilibrium calculations for multiphysics simulations performed using the open-source finite element / finite volume code Multiphysics Object Oriented Simulation Environment (MOOSE) [1]. While several methods have been proposed for accelerating phase equilibrium calculations [2], most focus on relatively small systems and often rely on a- priori knowledge of the state-space of the system. Nuclear materials, however, are often multi-component systems owing to the evolution of composition under irradiation and an approach based on a-priori mapping of phase diagram is often not enough. This work is aimed at demonstrating an on-the-fly surrogate modeling framework that uses active learning to reduce the number of full equilibrium calculations that must be performed. By combining with efficient coupling approaches, the surrogate framework helps in reducing the computational cost of thermodynamic equilibrium informed multiphysics simulations of nuclear materials. The performance is benchmarked against full coupling with the thermochemistry library Thermochimica [3]. This work uses a machine learning based approach for constructing surrogate models to predict the stable phases in a multicomponent system. The surrogates were constructed using neural networks and Gaussian process classification. In this work, we compare the relative performance of the two methods. We also demonstrate the use of caching previous calculations by interpolating the values from nearest neighbors. References [1] Lindsay, A.D., et al. "2.0 – MOOSE: Enabling massively parallel multiphysics simulation", SoftwareX, 20 (2022): 101202. [2] Roos, W.A. and Zietsman J.H. "Accelerating complex chemical equilibrium calculations – A Review", Calphad, 77 (2022): 102380. [3] Piro, M.H.A., et al. "The thermochemistry library Thermochimica", Computational Materials Science, 67 (2013): 266-272.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Physics-Based Machine Learning Methods for U-235 Forensics Signatures

Signatures of low-intensity U-235 sources have been recently studied by utilizing a variety of machine learning (ML) classifiers using features derived from gamma spectral measurements collectedunder structured campaigns. Several ML classifiers, such as ensemble of tress and classification trees, revealed misleadingly-optimistic training error due to over-fitting, and furthermore,their performance is not directly relatable to the physical properties due to their data-driven, opaque designs. We present a regression-based ML method that first estimates the inverse distanceto the source and then utilizes a threshold to infer its presence, by representing the background as a source located at an infinite distance. For the inverse distance estimation, we study the ensembleof trees and Gaussian process regression methods, and a hyper parameter auto-tuning and selection method that employs five regression estimators. These methods avoid the over-fittingobserved in several ML classifiers, while providing the classification error nearly comparable to them based on independent test data. Their error is directly related to estimates of the inversephysical distance to source, and the precision of error determines the seperability property that determines the false alarm and missed detection rates. The property of monotonic decrease of thesource strength with increasing detector distance combined with Poisson distribution of measurements is utilized to analytically validate these methods by deriving the generalization equations ofunderlying regression methods.

Rao, Nageswara↗
DOE OSTI2024

Physics-Based Machine Learning Methods for U-235 Forensics Signatures

Signatures of low-intensity U-235 sources have been recently studied by utilizing a variety of machine learning (ML) classifiers using features derived from gamma spectral measurements collected under structured campaigns. Several ML classifiers, such as ensemble of tress and classification trees, revealed misleadingly-optimistic training error due to over-fitting, and furthermore, their performance is not directly relatable to the physical properties due to their data-driven, opaque designs. We present a regression-based ML method that first estimates the inverse distance to the source and then utilizes a threshold to infer its presence, by representing the background as a source located at an infinite distance. For the inverse distance estimation, we study the ensemble of trees and Gaussian process regression methods, and a hyper parameter auto-tuning and selection method that employs five regression estimators. These methods avoid the over-fitting observed in several ML classifiers, while providing the classification error nearly comparable to them based on independent test data. Their error is directly related to estimates of the inverse physical distance to source, and the precision of error determines the seperability property that determines the false alarm and missed detection rates. The property of monotonic decrease of the source strength with increasing detector distance combined with Poisson distribution of measurements is utilized to analytically validate these methods by deriving the generalization equations of underlying regression methods.

Rao, Nageswara↗
DOE OSTI2024

Visualization and Quantification of Wind Induced Variability in Hydrogen Clouds Following Releases of Liquid Hydrogen: Preprint

Well characterized experimental data for consequence model validation is important in progressing the use of liquid hydrogen as an energy carrier. In 2019, the Health and Safety Executive (HSE) undertook a series of liquid hydrogen dispersion and combustion experiments as a part of the Pre-normative Research into the Safe Use of Liquid Hydrogen (PRESLHY) project. In partnership between the National Renewable Energy Laboratory (NREL) and HSE, time and spatially varying hydrogen concentration measurements were made in 25 dispersion experiments and 23 congested ignition experiments associated with PRESLHY WP3 and WP5, respectively. These measurements were undertaken using the hydrogen wide area monitoring system developed by NREL. During the 23 congested ignition experiments, high variability was observed in the measured explosion severity during experiments with similar initial conditions. This led to the conclusion that wind, including localized gusts, had a large influence on the dispersion of the hydrogen, and therefore the quantity of hydrogen that was present in the congested region of the explosions. Using the hydrogen concentration measurements taken immediately prior to ignition, the hydrogen clouds were visualized in an attempt to rationalize the variability in overpressure between the tests. Gaussian process regression was applied to quantify the variability of the measured hydrogen concentrations. This analysis could also be used to guide modifications in experimental designs for future research on hydrogen combustion behavior.

HSR&D↗
DOE OSTI2025

Uncertainty based Online Ensemble on Non-Stationary Data for Fusion Science

Machine Learning (ML) is poised to play a pivotal role in the development and operation of next-generation fusion devices. Fusion data shows non-stationary behavior due to drifts in the data. The drifts can arise from both experimental evolution and machine wear-and-tear. ML models assume stationary distribution and fail to maintain performance when encountered with non-stationary data streams.Online learning can be used to continuously adapt the models with new data as it is acquired. However, traditional online learning can suffer from short-term performance degradation, as ground truth are not available before making the prediction. To address this challenge, we propose uncertainty aware ensemble approach for online learning. We use Deep Gaussian Process Approximation (DGPA) technique for calibrated uncertainty estimation and use the uncertainty values to guide a meta-algorithm that produces predictions based on ensemble of learners. Moreover, DGPA also provides uncertainty estimation along with the predictions for decision makers. This paper demonstrates that the proposed method outperforms traditional online learning approach, and a naive ensemble without uncertainty guidance by about 7% and 6%, respectively, on B-coil deflection prediction at DIII-D Fusion Facility.

Rajput, Kishansingh [Thomas Jefferson National Acc↗