Search NASA⌕ Search

SEARCH · Search NASA

Results for “Bayesian Methods”

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 1,279 records · Page 71

Proxy-Based Bayesian Inversion Of Poroelastic Simulations To Interpret Strain Tensor Data Measured During Well Testing

The long runtimes of 3D poroelastic numerical simulators makes it impractical to interpret deformation datasets using many inversion schemes. Recent advances in instrumentation have made it possible to measure the strain tensor during well testing, but the lack of robust inversion methods is limiting the ability to interpret these data. We have developed an inversion workflow that reduces the number of computations required to complete a Bayesian inversion using DREAMzs. The workflow trains a KNN model using output from the poroelastic simulator, and then uses the KNN model as a proxy for the simulator during inversion. The workflow also includes a strategy for ensuring the results from the proxy model converge to the results from the simulator, ensuring the accuracy of the final results. An idealized example configured to represent a well test in a deep aquifer is used to verify that the workflow correctly identifies parameters and characterizes noise. Field data measured using strainmeters during an injection test at an oil reservoir in Oklahoma are used to evaluate performance with a real dataset. The workflow identified 265 history matching solutions out of 1240 total simulation runs (21% acceptance ratio), and the results are used to characterize posterior parameter distribution and evaluate the prediction uncertainty. This approach makes it feasible to invert strain data measured during well testing and this has the potential to improve the characterization of aquifers and reservoirs.

Roudini, Soheil↗

Bayesian characterization of uncertainties surrounding fluvial flood hazard estimates

Fluvial floods drive severe risk to riverine communities. There is strong evidence of increasing flood hazards in many regions around the world. The choice of methods and assumptions used in flood hazard estimates can impact the design of risk management strategies. In this study, we characterize the expected flood hazards conditioned on the uncertain model structures, model parameters, and prior distributions of the parameters. We construct a Bayesian framework for river stage return level estimation using a nonstationary statistical model that relies exclusively on the Indian Ocean Dipole Index. We show that ignoring uncertainties can lead to biased estimation of expected flood hazards. We find that the considered model parametric uncertainty is more influential than model structures and model priors. Our results highlight the importance of incorporating uncertainty in extreme flood stage estimates, and are of practical use for informing water infrastructure designs in a changing climate.

54 ENVIRONMENTAL SCIENCES↗

Bayesian Optimized Deep Ensemble for Uncertainty Quantification of Deep Neural Networks: a System Safety Case Study on Sodium Fast Reactor Thermal Stratification Modeling

Deep neural networks (DNNs) are increasingly important to scientific computing and engineering system simulations. Accurate uncertainty quantification (UQ) for DNNs is critical in safety-sensitive engineering domains. Traditional Deep Ensemble (DE) methods, while easy to implement, frequently suffer from poorly calibrated uncertainty estimates and limited predictive accuracy due to reliance on fixed architectures with varied weight initializations. To address these issues, we introduce a workflow that combines Bayesian Optimization (BO) and DE. The workflow is modular, scalable, and integrates parallel BO initialized with Sobol sequences to individually optimize the hyperparameters of each ensemble member. This method enhances ensemble diversity, improves predictive accuracy, and provides reliable uncertainty estimates. We evaluate the proposed BODE approach in a sodium fast reactor thermal stratification modeling case study, where we used a densely connected convolutional neural network to predict turbulent viscosity during the reactor transient with consideration of data noise. We benchmark its performance against several optimization approaches, including baseline deep ensemble, evolutionary algorithm-optimized ensemble, ensemble formed via random search combined with greedy selection, and a BO ensemble using random initialization. Here, our results demonstrate superior performance of the developed BODE approach. In noise-free scenarios, BODE notably reduces incorrect aleatoric uncertainty and significantly enhances predictive accuracy. Under conditions of 5% and 10% Gaussian noise, BODE adaptively quantifies uncertainty proportional to data noise, achieving up to an 80% reduction in root mean square error compared to baseline methods and producing well-calibrated prediction intervals.

Bayesian optimization↗

Soil moisture estimation using tomographic ground penetrating radar in a MCMC–Bayesian framework

In this paper, we focus on a hydrogeological inverse problem specifically targeting monitoring soil moisture variations using tomographic ground penetrating radar (GPR) travel time data. Technical challenges exist in the inversion of GPR tomographic data for handling non-uniqueness, nonlinearity and high-dimensionality of unknowns. We have developed a new method for estimating soil moisture fields from crosshole GPR data. It uses a pilot-point method to provide a low-dimensional representation of the relative dielectric permittivity field of the soil, which is the primary object of inference: the field can be converted to soil moisture using a petrophysical model. We integrate a multi-chain Markov chain Monte Carlo (MCMC)–Bayesian inversion framework with the pilot point concept, a curved-ray GPR travel time model, and a sequential Gaussian simulation algorithm, for estimating the dielectric permittivity at pilot point locations distributed within the tomogram, as well as the corresponding geostatistical parameters (i.e., spatial correlation range). We infer the dielectric permittivity as a probability density function, thus capturing the uncertainty in the inference. The multi-chain MCMC enables addressing high-dimensional inverse problems as required in the inversion setup. The method is scalable in terms of number of chains and processors, and is useful for computationally demanding Bayesian model calibration in scientific and engineering problems. The proposed inversion approach can successfully approximate the posterior density distributions of the pilot points, and capture the true values. The computational efficiency, accuracy, and convergence behaviors of the inversion approach were also systematically evaluated, by comparing the inversion results obtained with different levels of noises in the observations, increased observational data, as well as increased number of pilot points.

54 ENVIRONMENTAL SCIENCES↗

Adaptive Bayesian Inference for Prediction

Performing rigorous predictive simulation necessitates that all available data be incorporated into the modeling process in order to accurately quantify uncertainties in predictions of the model. Statistical Bayesian inference is the most commonly used approach for incorporating data into uncertainty estimates. Statistical Bayesian inference uses data and an assumed error model to inform posterior distributions of model input variables and model discrepancies. An explicit characterization of the posterior distribution is not formed, but rather methods such as Markov chain Monte-Carlo (MCMC) are used to draw samples from the posterior. These samples are then used to evaluate statistics, such as integrals, of predicted quantities of interest (QoI). Recently the authors have developed an alternative reformulation of Bayesian inference based upon a measure-theoretic inversion technique. Unlike traditional statistical Bayesian inference, this reformulation allows us to directly interrogate the posterior distribution and, consequently, to generate samples from the posterior without MCMC. In this work we use forward propagation uncertainty quantification techniques to reduce the cost of approximating the posterior distribution. Specifically we investigate goal-oriented adaptive sparse grid approximation methods. These adaptive methods concentrate high-fidelity model evaluations in regions of the parameter space that both significantly contribute to the uncertainty in specified QoI and are informed by the available data. Numerical results are presented to compare our approach with the standard statistical Bayesian approach and to demonstrate the effectiveness of the goal-oriented adaptivity.

97 MATHEMATICS AND COMPUTING↗

Variable Selection for High Dimensional Bayesian Density Estimation: Application to Human Exposure Simulation

Summary Numerous studies have linked ambient air pollution and adverse health outcomes. Many studies of this nature relate outdoor pollution levels measured at a few monitoring stations with health outcomes. Recently, computational methods have been developed to model the distribution of personal exposures, rather than ambient concentration, and then relate the exposure distribution to the health outcome. Although these methods show great promise, they are limited by the computational demands of the exposure model. We propose a method to alleviate these computational burdens with the eventual goal of implementing a national study of the health effects of air pollution exposure. Our approach is to develop a statistical emulator for the exposure model, i.e. we use Bayesian density estimation to predict the conditional exposure distribution as a function of several variables, such as temperature, human activity and physical characteristics of the pollutant. This poses a challenging statistical problem because there are many predictors of the exposure distribution and density estimation is notoriously difficult in high dimensions. To overcome this challenge, we use stochastic search variable selection to identify a subset of the variables that have more than just additive effects on the mean of the exposure distribution. We apply our method to emulate an ozone exposure model in Philadelphia.

Reich, Brian J.↗

Bayesian data fusion for spatial prediction of categorical variables in environmental sciences

First developed to predict continuous variables, Bayesian Maximum Entropy (BME) has become a complete framework in the context of space-time prediction since it has been extended to predict categorical variables and mixed random fields. This method proposes solutions to combine several sources of data whatever the nature of the information. However, the various attempts that were made for adapting the BME methodology to categorical variables and mixed random fields faced some limitations, as a high computational burden. The main objective of this paper is to overcome this limitation by generalizing the Bayesian Data Fusion (BDF) theoretical framework to categorical variables, which is somehow a simplification of the BME method through the convenient conditional independence hypothesis. The BDF methodology for categorical variables is first described and then applied to a practical case study: the estimation of soil drainage classes using a soil map and point observations in the sandy area of Flanders around the city of Mechelen (Belgium). The BDF approach is compared to BME along with more classical approaches, as Indicator CoKringing (ICK) and logistic regression. Estimators are compared using various indicators, namely the Percentage of Correctly Classified locations (PCC) and the Average Highest Probability (AHP). Although BDF methodology for categorical variables is somehow a simplification of BME approach, both methods lead to similar results and have strong advantages compared to ICK and logistic regression.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Integrating Survival Analysis with Bayesian Statistics to Forecast the Remaining Useful Life of a Centrifugal Pump Conditional to Multiple Fault Types

To improve the viability of nuclear power plants, there is a need to reduce their operational costs. Operational costs account for a significant portion of a plant’s yearly budget, due to their scheduled-based maintenance approach. In order to reduce these costs, proactive methods are required that estimate and forecast the state of a machine in real time to optimize maintenance schedules. In this research, we use Bayesian networks to develop a framework that can forecast the remaining useful life of a centrifugal pump. To do so, we integrate survival analysis with Bayesian statistics to forecast the health of the pump conditional to its current state. We complete our research by successfully using the Bayesian network on a case study. This solution provides an informed probabilistic viewpoint of the pumping system for the purpose of predictive maintenance.

42 ENGINEERING↗

A dynamic Bayesian optimized active recommender system for curiosity-driven partially Human-in-the-loop automated experiments

Optimization of experimental materials synthesis and characterization through active learning methods has been growing over the last decade, with examples ranging from measurements of diffraction on combinatorial alloys at synchrotrons, to searches through chemical space with automated synthesis robots for perovskites. In virtually all cases, the target property of interest for optimization is defined a priori with the ability to shift the trajectory of the optimization based on human-identified findings during the experiment is lacking. Thus, to highlight the best of both human operators and AI-driven experiments, here we present the development of a human–AI collaborated experimental workflow, via a Bayesian optimized active recommender system (BOARS), to shape targets on the fly with human real-time feedback. Here, the human guidance overpowers AI at early iteration when prior knowledge (uncertainty) is minimal (higher), while the AI overpowers the human during later iterations to accelerate the process with the human-assessed goal. We showcase examples of this framework applied to pre-acquired piezoresponse force spectroscopy of a ferroelectric thin film, and in real-time on an atomic force microscope, with human assessment to find symmetric hysteresis loops. It is found that such features appear more affected by subsurface defects than the local domain structure. This work shows the utility of human–AI approaches for curiosity driven exploration of systems across experimental domains.

36 MATERIALS SCIENCE↗

Learning an Eddy Viscosity Model Using Shrinkage and Bayesian Calibration: A Jet-in-Crossflow Case Study

In this paper, we demonstrate a statistical procedure for learning a high-order eddy viscosity model (EVM) from experimental data and using it to improve the predictive skill of a Reynolds-averaged Navier–Stokes (RANS) simulator. The method is tested in a three-dimensional (3D), transonic jet-in-crossflow (JIC) configuration. The process starts with a cubic eddy viscosity model (CEVM) developed for incompressible flows. It is fitted to limited experimental JIC data using shrinkage regression. The shrinkage process removes all the terms from the model, except an intercept, a linear term, and a quadratic one involving the square of the vorticity. The shrunk eddy viscosity model is implemented in an RANS simulator and calibrated, using vorticity measurements, to infer three parameters. The calibration is Bayesian and is solved using a Markov chain Monte Carlo (MCMC) method. A 3D probability density distribution for the inferred parameters is constructed, thus quantifying the uncertainty in the estimate. The phenomenal cost of using a 3D flow simulator inside an MCMC loop is mitigated by using surrogate models (“curve-fits”). A support vector machine classifier (SVMC) is used to impose our prior belief regarding parameter values, specifically to exclude nonphysical parameter combinations. The calibrated model is compared, in terms of its predictive skill, to simulations using uncalibrated linear and CEVMs. Finally, we find that the calibrated model, with one quadratic term, is more accurate than the uncalibrated simulator. The model is also checked at a flow condition at which the model was not calibrated.

42 ENGINEERING↗

Diffuse interstellar band at 8620 Å in rave: A new method for detecting the diffuse interstellar band in spectra of cool stars

Diffuse interstellar bands (DIBs) are usually observed in spectra of hot stars, where interstellar lines are rarely blended with stellar ones. The need for hot stars is a strong limitation in the number of sightlines we can observe and their distribution in the Galaxy, as hot stars are rare and concentrated in the Galactic plane. We are introducing a new method, where interstellar lines can be observed in spectra of cool stars in large spectroscopic surveys. The method is completely automated and does not require prior knowledge of the stellar parameters. The main step is a construction of the stellar spectrum, which is done by finding other observed spectra that lack interstellar features and are otherwise very similar to the spectrum in question. Such spectra are then combined into a single stellar spectrum template, matching the stellar component of the observed spectrum. We demonstrate the performance of this new method on a sample of 482,430 Radial Velocity Experiment survey spectra. However, many spectra have to be combined (48 on average) in order to achieve a signal-to-noise ratio high enough to measure the profile of the DIB at 8620 Å, hence limiting the spatial information about the interstellar medium. We compare its equivalent width with extinction maps and with Bayesian reddening, calculated for individual stars, and provide a linear relation between the equivalent width and reddening. Separately from the introduced method, we calculate equivalent widths of the DIB in spectra of hot stars with known extinction and compare all three linear relations.

79 ASTRONOMY AND ASTROPHYSICS↗

Multilevel sequential Monte Carlo samplers

Here, we study the approximation of expectations w.r.t. probability distributions associated to the solution of partial differential equations (PDEs); this scenario appears routinely in Bayesian inverse problems. In practice, one often has to solve the associated PDE numerically, using, for instance finite element methods and leading to a discretisation bias, with the step-size level h L . In addition, the expectation cannot be computed analytically and one often resorts to Monte Carlo methods. In the context of this problem, it is known that the introduction of the multilevel Monte Carlo (MLMC) method can reduce the amount of computational effort to estimate expectations, for a given level of error. This is achieved via a telescoping identity associated to a Monte Carlo approximation of a sequence of probability distributions with discretisation levels ${\infty}$ >h 0 >h 1 ...>h L . In many practical problems of interest, one cannot achieve an i.i.d. sampling of the associated sequence of probability distributions. A sequential Monte Carlo (SMC) version of the MLMC method is introduced to deal with this problem. In conclusion, it is shown that under appropriate assumptions, the attractive property of a reduction of the amount of computational effort to estimate expectations, for a given level of error, can be maintained within the SMC context.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Disaggregating measurement uncertainty from population variability and Bayesian treatment of uncensored results

In making low-level radioactivity measurements of populations, it is commonly observed that a substantial portion of net results are negative. Furthermore, the observed variance of the measurement results arises from a combination of measurement uncertainty and population variability. This paper presents a method for disaggregating measurement uncertainty from population variability to produce a probability density function (PDF) of possibly true results. To do this, simple, justifiable, and reasonable assumptions are made about the relationship of the measurements to the measurands (the 'true values'). The measurements are assumed to be unbiased, that is, that their average value is the average of the measurands. Using traditional estimates of each measurement's uncertainty to disaggregate population variability from measurement uncertainty, a PDF of measurands for the population is produced. Then, using Bayes's theorem, the same assumptions, and all the data from the population of individuals, a prior PDF is computed for each individual's measurand. These PDFs are non-negative, and their average is equal to the average of the measurement results for the population. The uncertainty in these Bayesian posterior PDFs is all Berkson with no remaining classical component. The methods are applied to baseline bioassay data from the Hanford site. The data include 90Sr urinalysis measurements on 128 people, 137Cs in vivo measurements on 5,337 people, and 239Pu urinalysis measurements on 3,270 people. The method produces excellent results for the 90Sr and 137Cs measurements, since there are nonzero concentrations of these global fallout radionuclides in people who have not been occupationally exposed. The method does not work for the 239Pu measurements in non-occupationally exposed people because the population average is essentially zero.

61 RADIATION PROTECTION AND DOSIMETRY↗

CATALOG MATCHING WITH ASTROMETRIC CORRECTION AND ITS APPLICATION TO THE HUBBLE LEGACY ARCHIVE

Object cross-identification in multiple observations is often complicated by the uncertainties in their astrometric calibration. Due to the lack of standard reference objects, an image with a small field of view can have significantly larger errors in its absolute positioning than the relative precision of the detected sources within. We present a new general solution for the relative astrometry that quickly refines the World Coordinate System of overlapping fields. The efficiency is obtained through the use of infinitesimal three-dimensional rotations on the celestial sphere, which do not involve trigonometric functions. They also enable an analytic solution to an important step in making the astrometric corrections. In cases with many overlapping images, the correct identification of detections that match together across different images is difficult to determine. We describe a new greedy Bayesian approach for selecting the best object matches across a large number of overlapping images. The methods are developed and demonstrated on the Hubble Legacy Archive, one of the most challenging data sets today. We describe a novel catalog compiled from many Hubble Space Telescope observations, where the detections are combined into a searchable collection of matches that link the individual detections. The matches provide descriptions of astronomical objects involving multiple wavelengths and epochs. High relative positional accuracy of objects is achieved across the Hubble images, often sub-pixel precision in the order of just a few milliarcseconds. The result is a reliable set of high-quality associations that are publicly available online.

79 ASTRONOMY AND ASTROPHYSICS↗

Distinguishing dark matter from unresolved point sources in the Inner Galaxy with photon statistics

Data from the Fermi Large Area Telescope suggests that there is an extended excess of GeV gamma-ray photons in the Inner Galaxy. Identifying potential astrophysical sources that contribute to this excess is an important step in verifying whether the signal originates from annihilating dark matter. In this paper, we focus on the potential contribution of unresolved point sources, such as millisecond pulsars (MSPs). We propose that the statistics of the photons—in particular, the flux probability density function (PDF) of the photon counts below the point-source detection threshold—can potentially distinguish between the dark-matter and point-source interpretations. We calculate the flux PDF via the method of generating functions for these two models of the excess. Working in the framework of Bayesian model comparison, we then demonstrate that the flux PDF can potentially provide evidence for an unresolved MSP-like point-source population.

79 ASTRONOMY AND ASTROPHYSICS↗

CODEX weak lensing: concentration of galaxy clusters at z ~ 0.5

Here, we present a stacked weak-lensing analysis of 27 richness selected galaxy clusters at 0.40 ≤ z ≤ 0.62 in the COnstrain Dark Energy with X-ray galaxy clusters (CODEX) survey. The fields were observed in five bands with the Canada–France–Hawaii Telescope (CFHT). We measure the stacked surface mass density profile with a 14σ significance in the radial range 0.1 < RMpch -1 < 2.5. The profile is well described by the halo model, with the main halo term following a Navarro–Frenk–White profile (NFW) profile and including the off-centring effect. We select the background sample using a conservative colour–magnitude method to reduce the potential systematic errors and contamination by cluster member galaxies. We perform a Bayesian analysis for the stacked profile and constrain the best-fitting NFW parameters M 200c =6.6$+1.0\atop{-0.8}$×10 14 h -1 M⊙ and c 200c =3.7$+0.7\atop{-0.6}$. The off-centring effect was modelled based on previous observational results found for redMaPPer Sloan Digital Sky Survey clusters. Our constraints on M200c and c200c allow us to investigate the consistency with numerical predictions and select a concentration–mass relation to describe the high richness CODEX sample. Comparing our best-fitting values for M200c and c200c with other observational surveys at different redshifts, we find no evidence for evolution in the concentration–mass relation, though it could be mitigated by particular selection functions. Similar to previous studies investigating the X-ray luminosity–mass relation, our data suggest a lower evolution than expected from self-similarity.

79 ASTRONOMY AND ASTROPHYSICS↗

Internship Final Report on the unsupervised learning sensor fusion (ULSF) approach

This paper describes a summer internship project undertaken at Sandia National Labs (SNL), both current status and future work. The project was to explore various machine learning approaches for use on turbulent flow data. Specifically, unsupervised classification of turbulent flow data was explored. First, the usage of models in this field is discussed, and several issues in the common usage of the models are identified. Solutions to these issues are then proposed, in the form of a Bayesian filtering approach which probabilistically incorporates multiple sources of data to improve confidence in a result. Several types of sensors are suggested for this method, the incorporation of which range from semi-supervised learning approaches to fully unsupervised. These approaches are then tested on several turbulent flow cases.

97 MATHEMATICS AND COMPUTING↗

What is ab initio in nuclear theory?

Ab initio has been used as a label in nuclear theory for over two decades. Its meaning has evolved and broadened over the years. We present our interpretation, briefly review its historical use, and discuss its present-day relation to theoretical uncertainty quantification.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗