Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bayesian Inference”

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 631 records · Page 35

Fast alogorithms for Bayesian uncertainty quantification in large-scale linear inverse problems based on low-rank partial Hessian approximations

We consider the problem of estimating the uncertainty in large-scale linear statistical inverse problems with high-dimensional parameter spaces within the framework of Bayesian inference. When the noise and prior probability densities are Gaussian, the solution to the inverse problem is also Gaussian, and is thus characterized by the mean and covariance matrix of the posterior probability density. Unfortunately, explicitly computing the posterior covariance matrix requires as many forward solutions as there are parameters, and is thus prohibitive when the forward problem is expensive and the parameter dimension is large. However, for many ill-posed inverse problems, the Hessian matrix of the data misfit term has a spectrum that collapses rapidly to zero. We present a fast method for computation of an approximation to the posterior covariance that exploits the lowrank structure of the preconditioned (by the prior covariance) Hessian of the data misfit. Analysis of an infinite-dimensional model convection-diffusion problem, and numerical experiments on large-scale 3D convection-diffusion inverse problems with up to 1.5 million parameters, demonstrate that the number of forward PDE solves required for an accurate low-rank approximation is independent of the problem dimension. This permits scalable estimation of the uncertainty in large-scale ill-posed linear inverse problems at a small multiple (independent of the problem dimension) of the cost of solving the forward problem.

97 MATHEMATICS AND COMPUTING↗

Uncertainty quantification in the presence of limited climate model data with discontinuities.

Uncertainty quantification in climate models is challenged by the prohibitive cost of a large number of model evaluations for sampling. Another feature that often prevents classical uncertainty analysis from being readily applicable is the bifurcative behavior in the climate data with respect to certain parameters. A typical example is the Meridional Overturning Circulation in the Atlantic Ocean. The maximum overturning stream function exhibits a discontinuity across a curve in the space of two uncertain parameters, namely climate sensitivity and CO2 forcing. In order to propagate uncertainties from model parameters to model output we use polynomial chaos (PC) expansions to represent the maximum overturning stream function in terms of the uncertain climate sensitivity and CO2 forcing parameters. Since the spectral methodology assumes a certain degree of smoothness, the presence of discontinuities suggests that separate PC expansions on each side of the discontinuity will lead to more accurate descriptions of the climate model output compared to global PC expansions. We propose a methodology that first finds a probabilistic description of the discontinuity given a number of data points. Assuming the discontinuity curve is a polynomial, the algorithm is based on Bayesian inference of its coefficients. Markov chain Monte Carlo sampling is used to obtain joint distributions for the polynomial coefficients, effectively parameterizing the distribution over all possible discontinuity curves. Next, we apply the Rosenblatt transformation to the irregular parameter domains on each side of the discontinuity. This transformation maps a space of uncertain parameters with specific probability distributions to a space of i.i.d standard random variables where orthogonal projections can be used to obtain PC coefficients. In particular, we use uniform random variables that are compatible with PC expansions based on Legendre polynomials. The Rosenblatt transformation and the corresponding PC expansions for the model output on either side of the discontinuity are applied successively for several realizations of the discontinuity curve. The climate model output and its associated uncertainty at specific design points is then computed by taking a quadrature-based integration average over PC expansions corresponding to possible realizations of the discontinuity curve.

54 ENVIRONMENTAL SCIENCES↗

Physics-based, Bayesian sequential detection method and system for radioactive contraband

A distributed sequential method and system for detecting and identifying radioactive contraband from highly uncertain (noisy) low-count, radionuclide measurements, i.e. an event mode sequence (EMS), using a statistical approach based on Bayesian inference and physics-model-based signal processing based on the representation of a radionuclide as a monoenergetic decomposition of monoenergetic sources. For a given photon event of the EMS, the appropriate monoenergy processing channel is determined using a confidence interval condition-based discriminator for the energy amplitude and interarrival time and parameter estimates are used to update a measured probability density function estimate for a target radionuclide. A sequential likelihood ratio test is then used to determine one of two threshold conditions signifying that the EMS is either identified as the target radionuclide or not, and if not, then repeating the process for the next sequential photon event of the EMS until one of the two threshold conditions is satisfied.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Update on a short-distance $D^0$-meson mixing calculation with $N_f=2+1$ flavors

We present an update on our calculation of the short-distance $D^0$-meson mixing hadronic matrix elements. The analysis is performed on the MILC collaboration's $N_f=2+1$ asqtad configurations. We use asqtad light valence quarks and the Sheikoleslami-Wohlert action with the Fermilab interpretation for the valence charm quark. SU(3), partially quenched, rooted, staggered heavy-meson chiral perturbation theory is used to extrapolate to the chiral-continuum limit. Systematic errors arising from the chiral-continuum extrapolation, heavy-quark discretization, and quark-mass uncertainties are folded into the statistical errors from the chiral-continuum fits with methods of Bayesian inference. A preliminary error budget for all five operators is presented.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Monte Carlo Global QCD Analyses of the Pion Parton Distribution Functions

As the lightest hadron, the pion presents itself as a dichotomy. While being the pseudo Goldstone boson associated with chiral symmetry breaking, it is simultaneously regarded as the lightest pseudoscalar meson typically composed of constituent quark and antiquark. As the chiral symmetry shares the same global symmetry as QCD, studying the pion and its internal structure is key in understanding one of the four fundamental forces of nature, namely the quantum chromodynamic strong interaction. We study the pion's parton distribution functions (PDFs), which are universal quantities that describe the structure of the pion in terms of its constituent quarks, antiquarks, and gluons. Our goal is to discover what the available data reveal for these universal quantities, namely the PDFs. Through the use of factorization theorems and perturbative QCD, we use Monte Carlo (MC) methods to extract the PDFs from the available Drell-Yan (DY) and leading neutron (LN) data. While the DY process involves two hadrons colliding, the LN electroproduction involves an electron beam incident on a target nucleon. We can constrain well the pion PDFs at large momentum fraction using DY data, and at low momentum fraction using LN data. We also use threshold resummation in the DY process to gather predictable higher order terms associated with the soft gluon radiation because their contributions to the cross section are nontrivial. In this dissertation, we parametrize the pion PDFs and fit those parameters to the available data. We make use of Bayesian inference and use MC techniques to sample the parameter space. Three sets of results are presented. We first extract the pion PDFs by fitting the DY and LN data. Then, we include transverse momentum dependent DY data in conjunction with the DY and LN data to extract the pion PDFs. Finally, we apply various methods of threshold resummation to the DY cross section and extract pion PDFs. The pion PDFs presented here are the first global QCD analyses performed as well as the first MC extracted pion PDFs, which are at the forefront of the pion PDF community.

Barry, Patrick↗

Statistical modeling support for calibration of a multiphysics model of subcooled boiling flows

Nuclear reactor system analyses rely on multiple complex models which describe the physics of reactor neutronics, thermal hydraulics, structural mechanics, coolant physico-chemistry, etc. Such coupled multiphysics models require extensive calibration and validation before they can be used in practical system safety study and/or design/technology optimization. This paper presents an application of statistical modeling and Bayesian inference in calibrating an example multiphysics model of subcooled boiling flows which is widely used in reactor thermal hydraulic analysis. The presence of complex coupling of physics in such a model together with the large number of model inputs, parameters and multidimensional outputs poses significant challenge to the model calibration method. However, the method proposed in this work is shown to be able to overcome these difficulties while allowing data (observation) uncertainty and model inadequacy to be taken into consideration. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Inverse Uncertainty Quantification of Reactor Simulation with Polynomial Chaos Surrogate Model

Reliable uncertainty quantification (UQ) and sensitivity analysis (SA) is significant part of modelling and simulation. This is particularly important in the nuclear community due to the fact that the licensing of nuclear installations has shifted from conservative to Best Estimate plus Uncertainty (BEPU) methodologies. UQ can be broadly classified as forward UQ and inverse UQ. Forward UQ is the process of quantifying the uncertainties in simulation outputs by propagating the uncertainties in input parameters. It requires knowledge in the input uncertainty, which is often specified by ad-hoc expert judgment which lacks mathematical rigor. The inverse UQ is the process to determine the uncertainty in input parameters given observations of the output. The purpose of this paper is to replace such ad-hoc expert judgment of the statistical properties of input model parameters in nuclear reactor simulation. This summary presents work on inverse UQ of a nonlinear Point Reactor Kinetics Equations - lumped thermal-hydraulics parameter coupled problem. Bayesian inference and Markov Chain Monte Carlo (MCMC) are used for the inverse UQ. Because MCMC requires a large number of simulations which is impractical for models that are computationally expensive, surrogate models is constructed by generalized Polynomial Chaos (PC) expansion technique. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Damage prognosis of adhesively-bonded joints in laminated composite structural components of unmanned aerial vehicles

The extensive use of lightweight advanced composite materials in unmanned aerial vehicles (UAVs) drastically increases the sensitivity to both fatigue- and impact-induced damage of their critical structural components (e.g., wings and tail stabilizers) during service life. The spar-to-skin adhesive joints are considered one of the most fatigue sensitive subcomponents of a lightweight UAV composite wing with damage progressively evolving from the wing root. This paper presents a comprehensive probabilistic methodology for predicting the remaining service life of adhesively-bonded joints in laminated composite structural components of UAVs. Non-destructive evaluation techniques and Bayesian inference are used to (i) assess the current state of damage of the system and, (ii) update the probability distribution of the damage extent at various locations. A probabilistic model for future loads and a mechanics-based damage model are then used to stochastically propagate damage through the joint. Combined local (e.g., exceedance of a critical damage size) and global (e.g.. flutter instability) failure criteria are finally used to compute the probability of component failure at future times. The applicability and the partial validation of the proposed methodology are then briefly discussed by analyzing the debonding propagation, along a pre-defined adhesive interface, in a simply supported laminated composite beam with solid rectangular cross section, subjected to a concentrated load applied at mid-span. A specially developed Eliler-Bernoulli beam finite element with interlaminar slip along the damageable interface is used in combination with a cohesive zone model to study the fatigue-induced degradation in the adhesive material. The preliminary numerical results presented are promising for the future validation of the methodology.

42 ENGINEERING↗

Using Bayesian analysis and Gaussian processes to infer electron temperature and density profiles on the Mega-Ampere Spherical Tokamak experiment

A unified, Bayesian inference of midplane electron temperature and density profiles using both Thomson scattering (TS) and interferometric data is presented. Beyond the Bayesian nature of the analysis, novel features of the inference are the use of a Gaussian process prior to infer a mollification length-scale of inferred profiles and the use of Gauss-Laguerre quadratures to directly calculate the depolarisation term associated with the TS forward model. Results are presented from an application of the method to data from the high resolution TS system on the Mega-Ampere Spherical Tokamak, along with a comparison to profiles coming from the standard analysis carried out on that system.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

IZI: INFERRING THE GAS PHASE METALLICITY (Z) AND IONIZATION PARAMETER (q) OF IONIZED NEBULAE USING BAYESIAN STATISTICS

We present a new method for inferring the metallicity (Z) and ionization parameter (q) of H II regions and star-forming galaxies using strong nebular emission lines (SELs). We use Bayesian inference to derive the joint and marginalized posterior probability density functions for Z and q given a set of observed line fluxes and an input photoionization model. Our approach allows the use of arbitrary sets of SELs and the inclusion of flux upper limits. The method provides a self-consistent way of determining the physical conditions of ionized nebulae that is not tied to the arbitrary choice of a particular SEL diagnostic and uses all the available information. Unlike theoretically calibrated SEL diagnostics, the method is flexible and not tied to a particular photoionization model. We describe our algorithm, validate it against other methods, and present a tool that implements it called IZI. Using a sample of nearby extragalactic H II regions, we assess the performance of commonly used SEL abundance diagnostics. We also use a sample of 22 local H II regions having both direct and recombination line (RL) oxygen abundance measurements in the literature to study discrepancies in the abundance scale between different methods. We find that oxygen abundances derived through Bayesian inference using currently available photoionization models in the literature can be in good (∼30%) agreement with RL abundances, although some models perform significantly better than others. We also confirm that abundances measured using the direct method are typically ∼0.2 dex lower than both RL and photoionization-model-based abundances.

79 ASTRONOMY AND ASTROPHYSICS↗

Accelerating parameter inference with graphics processing units

Gravitational wave Bayesian parameter inference involves repeated comparisons of GW data to generic candidate predictions. Even with algorithmically efficient methods like RIFT or reduced-order quadrature, the time needed to perform these calculations and overall computational cost can be significant compared to the minutes to hours needed to achieve the goals of low-latency multimessenger astronomy. By translating some elements of the RIFT algorithm to operate on graphics processing units (GPU), we demonstrate substantial performance improvements, enabling dramatically reduced overall cost and latency.

97 MATHEMATICS AND COMPUTING↗

Multiverse: Bayesian model selection for neural networks

Multiverse is a code repository for a set of tools for Bayesian model selection for neural networks. The goal of the tools is to provide capabilities for selecting among prior and model specifications for Bayesian neural networks. Multiverse will include tools for creating Bayesian neural networks, evaluating the Bayesian model evidence, and performing inference in Bayesian neural networks. These components are written in Python, a high-level programming language that takes advantage of the Python ecosystem of high-quality open-source packages for machine learning and signal processing.

Klein, Natalie↗

Lens Model Accuracy in the Expected LSST Lensed AGN Sample

Strong gravitational lensing of active galactic nuclei (AGN) enables measurements of cosmological parameters through time-delay cosmography (TDC). With data from the upcoming LSST survey, we anticipate using a sample of O(1000) lensed AGN for TDC. To prepare for this dataset and enable this measurement, we construct and analyze a realistic mock sample of 1300 systems drawn from the OM10 (Oguri & Marshall 2010) catalog of simulated lenses with AGN sources at $z<3.1$ in order to test a key aspect of the analysis pipeline, that of the lens modeling. We realize the lenses as power law elliptical mass distributions and simulate 5-year LSST i-band coadd images. From every image, we infer the lens mass model parameters using neural posterior estimation (NPE). Focusing on the key model parameters, $θ_E$ (the Einstein Radius) and $γ_{lens}$ (the projected mass density profile slope), with consistent mass-light ellipticity correlations in test and training data, we recover $θ_E$ with less than 1% bias per lens, 6.5% precision per lens and $γ_{lens}$ with less than 3% bias per lens, 8% precision per lens. We find that lens light subtraction prior to modeling is only useful when applied to data sampled from the training prior. If emulated deconvolution is applied to the data prior to modeling, precision improves across all parameters by a factor of 2. Finally, we combine the inferred lens mass models using Bayesian Hierarchical Inference to recover the global properties of the lens sample with less than 1% bias.

Venkatraman, Padmavathi [Illinois U., Urbana; KIPA↗