Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bayesian Robustness”

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 235 records · Page 13

Projective Integral Updates for High-Dimensional Variational Inference

Variational inference is an approximation framework for Bayesian inference that seeks to improve quantified uncertainty in predictions by optimizing a simplified distribution over parameters to stand in for the full posterior. Capturing model variations that remain consistent with training data enables more robust predictions by reducing parameter sensitivity. This work introduces a fixed-point optimization for variational inference that is applicable when every feasible log density can be expressed as a linear combination of functions from a given basis. In such cases, the optimizer becomes a fixed-point of projective integral updates. When the basis spans univariate quadratics in each parameter, the feasible distributions are Gaussian mean-fields and the projective integral updates yield quasi-Newton variational Bayes (QNVB). Other bases and updates are also possible. Since these updates require high-dimensional integration, this work begins by proposing an efficient quasirandom sequence of quadratures for mean-field distributions. Each iterate of the sequence contains two evaluation points that combine to correctly integrate all univariate quadratic functions and, if the mean-field factors are symmetric, all univariate cubics. More importantly, averaging results over short subsequences achieves periodic exactness on a much larger space of multivariate polynomials of quadratic total degree. The corresponding variational updates require four loss evaluations with standard (not second-order) backpropagation to eliminate error terms from over half of all multivariate quadratic basis functions. Furthermore, this integration technique is motivated by first proposing stochastic blocked mean-field quadratures, which may be useful in other contexts. A PyTorch implementation of QNVB allows for better control over model uncertainty during training than competing methods. Experiments demonstrate superior generalizability for multiple learning problems and architectures.

Gaussian mean-field↗

Density reconstruction in convergent high-energy-density systems using x-ray radiography and Bayesian inference

X-ray radiography is a technique frequently used to diagnose convergent high-energy-density (HED) systems, such as inertial confinement fusion implosion, and provides unique information that is not available through self-emission measurements. In this work, we investigate the scope and limits of that information using a radiography simulation combined with the Bayesian inference workflow. The accuracy of density reconstruction from simulated radiographs of spherical implosions driven with 27 kJ laser energy is assessed, including the increase or decrease in accuracy due to the addition of Lagrangian marker layers, Poisson noise, and improved prior information. This work is the first to present the full uncertainty distributions inferred from radiography analysis in HED systems and demonstrates the importance of constructing the full posterior probability density, as opposed to a point estimate, due to the modal structure of the likelihood surface introduced by typical experimental noise sources. This general methodology can be used both for robust analysis of radiographic data and for an improved design of radiography experiments by modeling the full experimental system.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Minimally Informative Prior Distributions for PSA

A salient feature of Bayesian inference is its ability to incorporate information from a variety of sources into the inference model, via the prior distribution (hereafter simply “the prior”). However, over-reliance on old information can lead to priors that dominate new data. Some analysts seek to avoid this by trying to work with a minimally informative prior distribution. Another reason for choosing a minimally informative prior is to avoid the often-voiced criticism of subjectivity in the choice of prior. Minimally informative priors fall into two broad classes: 1) so-called noninformative priors, which attempt to be completely objective, in that the posterior distribution is determined as completely as possible by the observed data, the most well known example in this class being the Jeffreys prior, and 2) priors that are diffuse over the region where the likelihood function is nonnegligible, but that incorporate some information about the parameters being estimated, such as a mean value. In this paper, we compare four approaches in the second class, with respect to their practical implications for Bayesian inference in Probabilistic Safety Assessment (PSA). The most commonly used such prior, the so-called constrained noninformative prior, is a special case of the maximum entropy prior. This is formulated as a conjugate distribution for the most commonly encountered aleatory models in PSA, and is correspondingly mathematically convenient; however, it has a relatively light tail and this can cause the posterior mean to be overly influenced by the prior in updates with sparse data. A more informative prior that is capable, in principle, of dealing more effectively with sparse data is a mixture of conjugate priors. A particular diffuse nonconjugate prior, the logistic-normal, is shown to behave similarly for some purposes. Finally, we review the so-called robust prior. Rather than relying on the mathematical abstraction of entropy, as does the constrained noninformative prior, the robust prior places a heavy-tailed Cauchy prior on the canonical parameter of the aleatory model.

99 GENERAL AND MISCELLANEOUS↗

Probabilistic Mixture Model-Based Spectral Unmixing

Spectral unmixing attempts to decompose a spectral ensemble into the constituent pure spectral signatures (called endmembers) along with the proportion of each endmember. This is essential for techniques like hyperspectral imaging (HSI) used in environment monitoring, geological exploration, etc. Several spectral unmixing approaches have been proposed, many of which are connected to hyperspectral imaging. However, most extant approaches assume highly diverse collections of mixtures and extremely low-loss spectroscopic measurements. Additionally, current non-Bayesian frameworks do not incorporate the uncertainty inherent in unmixing. We propose a probabilistic inference algorithm that explicitly incorporates noise and uncertainty, enabling us to unmix endmembers in collections of mixtures with limited diversity. We use a Bayesian mixture model to jointly extract endmember spectra and mixing parameters while explicitly modeling observation noise and the resulting inference uncertainties. We obtain approximate distributions over endmember coordinates for each set of observed spectra while remaining robust to inference biases from the lack of pure observations and the presence of non-isotropic Gaussian noise. As a direct impact of our methodology, access to reliable uncertainties on the unmixing solutions would enable robust solutions to noise, as well as informed decision-making for HSI applications and other unmixing problems.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A Bayesian inferencing framework for ultrasound wave speed measurements in metal additive manufacturing

Process-related changes during metal additive manufacturing introduce microstructural variability in the material properties of printed parts, directly affecting component reliability. Accurate estimation of these property variations with part performance are essential for quality assurance. Ultrasound testing offers a non-destructive means to estimate mechanical properties and detect defects; however, conventional analysis methods often neglect the influence of microstructural variability, limiting their effectiveness. Here, this research presents a Bayesian inference technique for quantifying wave speed uncertainty from ultrasound measurements of metal additive manufactured parts. By integrating prior ultrasound data with a Bayesian model, the proposed approach generates posterior density estimates of wave speed that systematically account for manufacturing-induced variability and uncertainty. The novelty of this research lies in applying a Bayesian framework to analyze experimental ultrasound measurements within the context of metal additive manufacturing variability. The method enhances the accuracy of wave speed estimation by 64%, defect position by 50% and increases confidence associated with wave speed variance by 30% across different porosity levels, thereby providing a robust foundation for improved decision-making and increased reliability in additively manufactured components.

Additive manufacturing↗

Optimal Pulse Processing, Pile-Up Decomposition, and Applications of Silicon Drift Detectors at LCLS

Silicon drift detectors (SDDs) revolutionized spectroscopy in fields as diverse as geology and dentistry. For a subset of experiments at ultrafast, X-ray free-electron lasers (FELs), SDDs can make substantial contributions. Often the unknown spectrum is interesting, carrying science data, or the background measurement is useful to identify unexpected signals. Many measurements involve only several discrete photon energies known a priori, allowing single-event decomposition of pile-up and spectroscopic photon counting. We designed a pulse function and demonstrated that the signal amplitude (i.e., proportional to the detected energy and obtained from fitting with the pulse function), rise time, and pulse height are interrelated, and at short peaking times, the pulse height and pulse area are not optimal estimators for detected energy; instead, the signal amplitude and rise time are obtained for each pulse by fitting, thus removing the need for pulse shaping. By avoiding pulse shaping, rise times of tens of nanoseconds resulted in reduced pulse pile-up and allowed decomposition of remaining pulse pile-up at photon separation times down to hundreds of nanoseconds while yielding time-of-arrival information with the precision of 10 ns. Waveform fitting yields simultaneously high energy resolution and high counting rates (two orders of magnitude higher than current digital pulse processors). At pulsed sources or high photon rates, photon pile-up still occurs. We showed that pile-up spectrum fitting is relatively simple and preferable to pile-up spectrum deconvolution. We then developed a photon pile-up statistical model for constant intensity sources, extended it to variable intensity sources (typical for FELs), and used it to fit a complex pileup spectrum. We subsequently developed a Bayesian pile-up decomposition method that allows decomposing pile-up of single events with up to six photons from six monochromatic lines with 99% accuracy. The usefulness of SDDs will continue into the X-ray FEL era of science. Their successors, the ePixS hybrid pixel detectors, already offer hundreds of pixels, each with a similar performance to an SDD, in a compact, robust and affordable package.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Bayesian Design of Experiments for Logistic Regression to Evaluate Multiple Forensic Algorithms

When evaluating the performance of several forensic classification algorithms, it is desirable to construct a design that considers a variety of performance levels for each of the algorithms. We describe a strategy to use Bayesian design of experiments with multiple prior estimates to capture anticipated performance. Our goal is to characterize results from the different classification algorithms as a function of multiple explanatory variables and use this to choose a design about which units to test. Bayesian design of experiments has been successful for generalized linear models, including logistic regression models. Here, we develop methodology for the case where there are several potentially nonoverlapping priors for anticipated performance under consideration. The weighted priors method performs well for a broad range of true underlying model parameter choices and is more robust when compared to other candidate design choices. Additionally, we show how this can be applied in the multivariate input case and provide some useful summary measures. The shared information plot is used to evaluate design point allocation, and the D-value difference plot allows for the comparison of design performance across multiple potential parameter values in higher dimensions. Here, we illustrate the methods with several examples.

97 MATHEMATICS AND COMPUTING↗

Exclusion limits from data of directional dark matter detectors

Directional detection is a promising search strategy to discover Galactic dark matter. Taking advantage on the rotation of the Solar System around the Galactic center through the dark matter halo, it allows us to show a direction dependence of weakly interacting massive particle (WIMP) events. Even though the goal of directional search is to identify a WIMP positive detection, exclusion limits are still needed for very low exposure with a rather large background contamination, such as the one obtained with prototype experiments. Data of directional detectors are composed of energy and a 3D track for all recoiling nuclei. However, to set robust exclusion limits, we focus on the angular part of the event distribution, arguing that the energy part of the background distribution is unknown. As the angular distributions of both background and WIMP events are known, a Bayesian approach to set exclusion limits is possible. In this paper, a new statistical method based on an extended likelihood is proposed, compared to existing ones, and is shown to be optimal. Eventually, a comprehensive study of the effect of a detector configuration on exclusion limits is presented. It includes the effect of having or not sense recognition, a finite angular resolution, taking into account energy threshold as well as some astrophysical uncertainties.

79 ASTRONOMY AND ASTROPHYSICS↗

Sparsifying priors for Bayesian uncertainty quantification in model discovery

We propose a probabilistic model discovery method for identifying ordinary differential equations governing the dynamics of observed multivariate data. Our method is based on the sparse identification of nonlinear dynamics (SINDy) framework, where models are expressed as sparse linear combinations of pre-specified candidate functions. Promoting parsimony through sparsity leads to interpretable models that generalize to unknown data. Instead of targeting point estimates of the SINDy coefficients, we estimate these coefficients via sparse Bayesian inference. The resulting method, uncertainty quantification SINDy (UQ-SINDy), quantifies not only the uncertainty in the values of the SINDy coefficients due to observation errors and limited data, but also the probability of inclusion of each candidate function in the linear combination. UQ-SINDy promotes robustness against observation noise and limited data, interpretability (in terms of model selection and inclusion probabilities) and generalization capacity for out-of-sample forecast. Sparse inference for UQ-SINDy employs Markov chain Monte Carlo, and we explore two sparsifying priors: the spike and slab prior, and the regularized horseshoe prior. UQ-SINDy is shown to discover accurate models in the presence of noise and with orders-of-magnitude less data than current model discovery methods, thus providing a transformative method for real-world applications which have limited data.

97 MATHEMATICS AND COMPUTING↗

Dwarf spheroidal J-factor likelihoods for generalized NFW profiles

ABSTRACT Indirect detection strategies of particle dark matter (DM) in Dwarf spheroidal satellite galaxies (dSphs) typically entail searching for annihilation signals above the astrophysical background. To robustly compare model predictions with the observed fluxes of product particles, most analyses of astrophysical data – which are generally frequentist – rely on estimating the abundance of DM by calculating the so-called J factor. This quantity is usually inferred from the kinematic properties of the stellar population of a dSph using the Jeans equation, commonly by means of Bayesian techniques that entail the presence (and additional systematic uncertainty) of prior choice. Here, extending earlier work, we develop a scheme to derive the profile likelihood for J factors of dwarf spheroidals for models with five or more free parameters. We validate our method on a publicly available simulation suite, released by the Gaia Challenge, finding satisfactory statistical properties for bias and probability coverage. We present the profile likelihood function and maximum likelihood estimates for the J-factor of 10 dSphs. As an illustration, we apply these profile likelihoods to recently published analyses of γ-ray data with the Fermi Large Area Telescope to derive new, consistent upper limits on the DM annihilation cross-section. We do this for a subset of systems, generally referred to as classical dwarfs. The implications of these findings for DM searches are discussed, together with future improvements and extensions of this technique.

79 ASTRONOMY AND ASTROPHYSICS↗

Bayesian quantification of observability and equation of state of twin stars

The possibility of discovering twin stars, two neutron stars (NSs) with the same mass but different radii, is usually studied in forward modelings by using a restricted number of NS matter equations of state (EOSs) encapsulating a first-order phase transition from hadronic to quark matter (QM). Informing our likelihood function with the NS radius data from GW170817 and using a metamodel with nine parameters capable of mimicking most NS EOSs available in the literature, we conduct a Bayesian quantification of the observability and underlying EOSs of twin stars. Of the accepted EOSs, between 12 and 18% yield twin stars, depending on the restrictions we place on the second branch. The possibility of twin stars remains robust even under recent observational constraints. Here, we show that many of these twin star scenarios are observable with currently available levels of accuracy in measuring NS radii. We also present the marginalized posterior probability density functions (PDFs) of every EOS parameter for each of four mass-radius correlation topologies. We find that the inferred EOS depends sensitively on not only whether twin stars are present, but also the category of twin stars, indicating that the observation of twin stars would provide a strong constraint on the underlying EOS. In particular, for two coexisting hybrid stars having QM cores at different densities, the PDF for QM speed of sound squared 𝑐$^2_{qm}$ has two peaks, one below and another above the conformal limit 𝑐$^2_{qm}$ = 1/3 predicted by perturbative QCD.

QCD in nuclear reactions↗

Modeling the Radio Foreground for Detection of CMB Spectral Distortions from the Cosmic Dawn and the Epoch of Reionization

Cosmic baryon evolution during the Cosmic Dawn and Reionization results in redshifted 21-cm spectral distortions in the cosmic microwave background (CMB). These encode information about the nature and timing of first sources over redshifts 30–6 and appear at meter wavelengths as a tiny CMB distortion along with the Galactic and extragalactic radio sky, which is orders of magnitude brighter. Therefore, detection requires precise methods to model foregrounds. We present a method of foreground fitting using maximally smooth (MS) functions. We demonstrate the usefulness of MS functions over traditionally used polynomials to separate foregrounds from the Epoch of Reionization (EoR) signal. We also examine the level of spectral complexity in plausible foregrounds using GMOSS, a physically motivated model of the radio sky, and find that they are indeed smooth and can be modeled by MS functions to levels sufficient to discern the vanilla model of the EoR signal. We show that MS functions are loss resistant and robustly preserve EoR signal strength and turning points in the residuals. Finally, we demonstrate that in using a well-calibrated spectral radiometer and modeling foregrounds with MS functions, the global EoR signal can be detected with a Bayesian approach with 90% confidence in 10 minutes’ integration.

79 ASTRONOMY AND ASTROPHYSICS↗

Probabilistic Groundwater Modeling of the RDX Plume at Los Alamos National Laboratory to Support Risk Assessment - 20359

A plume of the contaminant hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX) with concentrations greater than the New Mexico tap water drinking standard (9.66 ppb) is present in the regional aquifer near the southwestern boundary of Los Alamos National Laboratory (LANL). A risk assessment is performed for exposure to regional aquifer groundwater, with long-term predictions of RDX concentrations provided by a calibrated, probabilistic, numerical fate and transport model. The structure of the model is hierarchical, with the RDX Regional Aquifer (RA) groundwater model acting as the primary tool for analysis of downgradient RDX concentrations. The RA model is deeply informed by the conceptual site model (CSM) and is calibrated using site RDX concentration data and hydraulic head measurements, along with other analyses. The model is calibrated using data through December 2019. Where data are scarce other lines of evidence are used to inform inputs, including the multiphase RDX Vadose Zone (VZ) model and Pipe and Disk analytical screening tool. Model inputs are described with informative prior distributions using a robust approach to development that incorporates all available lines of evidence for every parameter used as an input to the RA model. Calibration is performed using a classical nonlinear optimization routine, which is then used to initialize a Bayesian calibration. The Bayesian calibration constrains the uncertainty in the classical calibration, ultimately providing posterior distributions for all model parameters. The challenges of the calibration include high-dimensional parameter space, including spatially heterogeneous hydraulic conductivities, comparatively sparse data, and low RDX concentrations. Posterior parameter distributions developed in the calibration process are then used for stochastic predictive model runs into the future. The result of the forward runs is spatially and temporally explicit estimates of head and concentration with uncertainty at all points in space and time. The probabilistic modeling approach presented here includes innovative computational and statistical methods that leverage high-performance computing (HPC) resources. It makes use of prior modeling work performed at the chromium plume site in the central LANL area, with extensive updates. The risk assessment will ultimately be used to support decision-making at the site, using multiple appropriately weighted sources of information, as well as uncertainty, sensitivity, and value-of-information analyses. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Joint state-parameter estimation for the reduced fracture model via the united filter

Here, in this paper, we introduce an effective United Filter method for jointly estimating the solution state and physical parameters in flow and transport problems within fractured porous media. Fluid flow and transport in fractured porous media are critical in subsurface hydrology, geophysics, and reservoir geomechanics. Reduced fracture models, which represent fractures as lower-dimensional interfaces, enable efficient multi-scale simulations. However, reduced fracture models also face accuracy challenges due to modeling errors and uncertainties in physical parameters such as permeability and fracture geometry. To address these challenges, we propose a United Filter method, which integrates the Ensemble Score Filter (EnSF) for state estimation with the Direct Filter for parameter estimation. EnSF, based on a score-based diffusion model framework, produces ensemble representations of the state distribution without deep learning. Meanwhile, the Direct Filter, a recursive Bayesian inference method, estimates parameters directly from state observations. The United Filter combines these methods iteratively: EnSF estimates are used to refine parameter values, which are then fed back to improve state estimation. Numerical experiments demonstrate that the United Filter method surpasses the state-of-the-art Augmented Ensemble Kalman Filter, delivering more accurate state and parameter estimation for reduced fracture models. This framework also provides a robust and efficient solution for PDE-constrained inverse problems with uncertainties and sparse observations.

Bayesian inference↗

Global fits of the cMSSM including the first LHC and XENON100 data

We present updated global fits of the constrained Minimal Supersymmetric Standard Model (cMSSM), including the most recent constraints from the ATLAS and CMS detectors at the LHC, as well as the most recent results of the XENON100 experiment. Our robust analysis takes into account both astrophysical and hadronic uncertainties that enter in the calculation of the rate of WIMP-induced recoils in direct detection experiment. We study the consequences for neutralino Dark Matter, and show that current direct detection data already allow to robustly rule out the so-called Focus Point region, therefore demonstrating the importance of particle astrophysics experiments in constraining extensions of the Standard Model of Particle Physics. We also observe an increased compatibility between results obtained from a Bayesian and a Frequentist statistical perspective. We find that upcoming ton-scale direct detection experiments will probe essentially the entire currently favoured region (at the 99% level), almost independently of the statistical approach used. Prospects for indirect detection of the cMSSM are further reduced.

79 ASTRONOMY AND ASTROPHYSICS↗

The Ages and Masses of a Million Galactic-disk Main-sequence Turnoff and Subgiant Stars from the LAMOST Galactic Spectroscopic Surveys

We present estimates of stellar age and mass for 0.93 million Galactic-disk main-sequence turnoff and subgiant stars from the LAMOST Galactic Spectroscopic Surveys. The ages and masses are determined by matching with stellar isochrones using a Bayesian algorithm, utilizing effective temperature T{sub eff}, absolute magnitude M{sub V}, metallicity [Fe/H], and α-element to iron abundance ratio [α/Fe] deduced from the LAMOST spectra. Extensive examinations suggest the age and mass estimates are robust. Overall, the sample stars have a median error of 34% for the age estimates, and half of the stars older than 2 Gyr have age uncertainties of only 20%–30%. The median error for the mass estimates of the whole sample of stars is ∼8%. The huge data set demonstrates good correlations among stellar age, [Fe/H] ([α/H]), and [α/Fe]. Particularly, double-sequence features are revealed in both the age–[α/Fe] and age–[Fe/H]([α/H]) spaces. In the [Fe/H]–[α/Fe] space, stars of 8–10 Gyr exhibit both the thin and thick disk sequences, while younger (older) stars show only the thin (thick) disk sequence, indicating that the thin disk became prominent 8–10 Gyr ago, while the thick disk formed earlier and was almost quenched 8 Gyr ago. Stellar ages exhibit positive vertical and negative radial gradients across the disk, and the outer disk of R ≳ 9 kpc exhibits a strong flare in stellar age distribution.

79 ASTRONOMY AND ASTROPHYSICS↗

Bayesian inference for high confidence signal validation and sensor calibration assessment - 328

Online Monitoring (OLM) systems has been investigated for applications in the nuclear industry for the past two decades. These systems offer economic and technological advantages over the currently implemented time based testing methods for sensors and associated instrumentations. The advantages offered not only augment economic viability of current nuclear reactors but will be an inevitable change called for to suit the technologies of the advanced reactors (Gen III+). Irrespective, the acceptance of OLM into the regulatory structure in the U.S. is still hindered by technical and structural clarity on uncertainty quantification used by these systems. A Bayesian inference technique is implemented in the form of a residual model to quantify uncertainty in OLM predictions. The developed system used Gaussian processes to represent the OLM predictions and model inadequacy, the stationary representation of model based uncertainty. In addition, model selection techniques are used in conjunction with numerical methods such as Genetic Algorithm (GA) and Gibb's sampler to add robustness to the estimation model inadequacy, particularly in the presence of drifts and faulted conditions. The prediction model developed herein find applicability in fault detection, obtaining real values of processes after the onset of faults etc. Further development looks into the extension of current method for a multiple output scale implementation. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗