Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bayesian model calibration”

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 361 records · Page 20

"Sintering" Models and In-Situ Experiments: Data Assimilation for Microstructure Prediction in SLS Additive Manufacturing of Nylon Components

Selective laser sintering methods are workhorses for additively manufacturing polymer-based components. The ease of rapid prototyping also means it is easy to produce illicit components. It is necessary to have a data-calibrated in-situ physical model of the build process in order to predict expected and defective microstructure characteristics that inform component provenance. Toward this end, sintering models are calibrated and characteristics such as component defects are explored. This is accomplished by assimilating multiple data streams, imaging analysis, and computational model predictions in an adaptive Bayesian parameter estimation algorithm. From these data sources, along with a phase-field model, bulk porosity distributions are inferred. Model parameters are constrained to physically-relevant search directions by sensitivity analysis, and then matched to predictions using adaptive sampling. Using this feedback loop, data-constrained estimates of sintering model parameters along with uncertainty bounds are obtained.

3D printing, Additive Manufacturing, polymer, sint↗

A Prefire Approach for Probabilistic Assessments of Postfire Debris-Flow Inundation

Increases in wildfire activity and rainfall intensification are driving more postfire debris flows (PFDF) in many regions around the world. PFDFs are most common in the first postfire year and may even occur before a fire is fully controlled. This underscores the importance of assessing postfire hazards before a fire starts. Evaluation of PFDF hazards prior to fire can help strategize interventions lessening the negative effects of future fires. However, debris-flow runout and inundation analyses are not routine in PFDF hazard assessments, partially due to time constraints and substantial uncertainties in boundary conditions. Here, we propose a prefire PFDF inundation assessment framework using a debris-flow runout model based on the Herschel-Bulkley (HB) rheology (HEC-RAS v6.1). We constrain model inputs and parameters using Bayesian posterior analysis, rainfall-runoff simulations, and a debris-flow volume model. We use observations from recent PFDF incidents in northern Arizona, USA, to calibrate model components and then apply our prefire inundation assessment framework in a nearby unburned area. Specifically, we (a) identify yield stress as the most influential factor on inundation extent and arrival time in a HB model, (b) establish posterior distributions for model parameters suitable for forward modeling by leveraging uncertainties in field observations, and (c) implement a predictive forward analysis in an area that has not burned recently to evaluate PFDF inundation under several future fire scenarios. This study improves our ability to assess postfire debris-flow hazards before a fire begins and provides guidance for future applications of single-phase rheological models when assessing PFDF hazards.

54 ENVIRONMENTAL SCIENCES↗

Determination of Quark-Gluon-Plasma Parameters from a Global Bayesian Analysis

The quality of data taken at RHIC and LHC as well as the success and sophistication of computational models for the description of ultra-relativistic heavy-ion collisions have advanced to a level that allows for the quantitative extraction of the transport properties of the Quark-Gluon-Plasma. However, the complexity of this task as well as the computational effort associated with it can only be overcome by developing novel methodologies: in this paper we outline such an analysis based on Bayesian Statistics and systematically compare an event-by-event heavy-ion collision model to data from the Large Hadron Collider. We simultaneously probe multiple model parameters including fundamental quark-gluon plasma properties such as the temperature-dependence of the specific shear viscosity η/s, calibrate the model to optimally reproduce experimental data, and extract quantitative constraints for all parameters simultaneously. As a result, the method is universal and easily extensible to other data and collision models.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

The Lifecycle of Bayesian Network Models Developed for Multi-Source Signature Assessment of Nuclear Programs

The Multi-Source Signatures for Nuclear Programs project, part of Pacific Northwest National Laboratory’s (PNNL) Signature Discovery Initiative, seeks to computationally capture expert assessment of multi-type information such as text, sensor output, imagery, or audio/video files, to assess nuclear activities through a series of Bayesian network (BN) models. These models incorporate knowledge from a diverse range of information sources in order to help assess a country’s nuclear activities. The models span engineering topic areas, state-level indicators, and facility-specific characteristics. To illustrate the development, calibration, and use of BN models for multi-source assessment, we present a model that predicts a country’s likelihood to participate in the international nuclear nonproliferation regime. We validate this model by examining the extent to which the model assists non-experts arrive at conclusions similar to those provided by nuclear proliferation experts. We also describe the PNNL-developed software used throughout the lifecycle of the Bayesian network model development.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

Bayesian Approach to Estimation of Water Table Elevations Using Historical Rasters as Prior Information 2019 - 20430

In cases of complex but only partially known geology and a lack of spatial control in observation well locations, water table elevation estimation is very challenging. In some cases, auxiliary information, such as observations of the movement of tracers, operation of injection and extraction wells, and calibration of groundwater models against the historical elevation data, can be combined with expert judgement to estimate flow directions in areas of sparse data and to aid in the production of more reliable contour maps (and associated rasters) than could be produced by relying on sparse well elevation data alone. Given a historical sequence of these raster maps, the question arises how to automate, to the extent possible, the process of producing new raster maps to reflect data from previous times, the current data and the operation of expert judgement. One solution is to adopt a Bayesian point of view and to regard the historical well elevation data, auxiliary information and historical raster maps as prior information. The well elevations for water table wells, as well as those for injection/extraction wells and the data associated with other relevant variables, can be viewed as predictors for the raster surface. From this prior information, we can, conditional on the values of these predictors for a new time period, compute an expected value map and a standard deviation map for the new raster. These then can be taken to specify a prior predictive distribution for the pixels in the new raster map. Then we condition the pixels, corresponding to water level observation wells within the raster, on the observed values in those wells (which in general will differ from the regression estimate) for the new time period. Given the smoothness of the water table surface, we then smooth the surface of deviations from the mean surface, based on the variograms of the historical rasters, and add this smoothed surface to the regression mean surface. The error structure of the produced raster map is defined by the regression error structure and the error due to smoothing based on the estimated variograms. This methodology has been developed and is being further refined for groundwater monitoring and remediation at LANL. It is a very flexible method that can also be applied with a variety of other predictors applied to model the water level wells in the area of interest over the historical record. The smoothness of the spatial process and its possible evolution over time can then be estimated from the residuals from this regression. This can be augmented by expert hydrogeological opinion based on site topography and hydrogeology. (authors)

54 ENVIRONMENTAL SCIENCES↗

Partially Collapsed Gibbs Samplers for Linear Inverse Problems and Applications to X-ray Imaging

Deblurring is one of the most common linear inverse problems in imaging, and most methods for deblurring an image assume that the convolution kernel, the so-called point spread function (PSF), is known or attempt to reconstruct the PSF simultaneously with the image. For many experimental imaging systems, like high energy X-ray radiography systems, the PSF must be computed from calibration imagery. The associated inverse problem is large-scale, so, in order to compute meaningful uncertainties with the reconstruction, it is common to use hierarchical Bayesian models and Markov chain Monte Carlo methods for sampling from the posterior. Classical Gibbs samplers can suffer from long autocorrelation lengths in the Markov chain when the reconstructed quantity in the inverse problem is finely discretized, resulting in poor computational efficiency of the Markov chain. In this work we present partially collapsed Gibbs samplers for linear inverse problems that greatly decrease correlations in the Markov chain, thus increasing the computational efficiency of the algorithms, and we present the results of the enhanced samplers on the problem of reconstructing the point spread function for an X-ray imaging system at the U.S. Department of Energy's Nevada National Security Site.

97 MATHEMATICS AND COMPUTING↗

Bayesian discovery of optimal reduced order models from mechanistic and experimental data: A case study of Pd penetration in TRISO fuels using BISON

TRistructural ISOtropic (TRISO) particles rely on a silicon carbide (SiC) layer as the primary structural material and barrier to metallic fission products (FPs) release. Accurate prediction of palladium (Pd) transport and penetration is therefore critical for qualifying TRISO fuels for advanced reactors. The empirical correlation for Pd penetration in BISON is derived from historical particle-fuel data, but cannot explain the large scatter in the experimental data that arises from varying experimental conditions. To aid fuel qualification, we previously developed a mechanistic reduced order model (ROM) using BISON that resolves these dependencies. Here, in this work we build on that mechanistic ROM and perform validation and quantify its uncertainty using Bayesian uncertainty quantification (UQ). calibration against a suite of in-pile and out-of-pile experiments spanning particle compositions, geometries, and operating conditions, and we benchmark it against the empirical correlation. Bayesian UQ identifies influential parameters, calibrates them to data, and yields predictive intervals. Results show that while the empirical correlation can be tuned to fit a single experiment type, it transfers poorly; the mechanistic ROM sustains accuracy with credible uncertainty across disparate conditions. This demonstrates a practical path—via Bayesian UQ applied to mechanistic ROMs—to leverage single-effect experiments for inferring in-reactor behavior and supporting TRISO fuel qualification.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Optimal sizing of battery energy storage systems for peak shaving and demand response using a degradation-aware Bayesian Optimization-Mixed-Integer Linear Programming framework

The increasing integration of renewable energy and rising electricity demand highlight the importance of battery energy storage systems for peak shaving and demand response. Unlike prior approaches that overlook operational impacts on degradation, this study proposes a Bayesian Optimization–Mixed Integer Linear Programming framework for optimal battery energy storage system sizing. In this framework, Mixed Integer Linear Programming determines short-term scheduling while a calibrated electrochemical model iteratively evaluates degradation. The central hypothesis is that the framework can efficiently identify optimal sizes that yield realistic and economically robust outcomes. The method is tested across three scenarios: peak shaving, peak shaving with energy-reduction demand response, and peak shaving with power-reduction demand response. Results show that the framework converge to the optimum within 20 iterations out of 150 possible sizes. Under baseline conditions, the framework consistently selects the smallest feasible system, minimizing unnecessary degradation costs from oversized storage. Sensitivity analyses reveal that larger systems are favored as demand rates or incentives increase. Comparisons of demand response programs indicate that power-reduction demand response offers greater economic benefits than energy-reduction demand response, although demand savings from peak shaving remain the dominant contributor to overall performance. This study demonstrates that the proposed framework balances computational tractability with degradation fidelity, identifies critical economic thresholds for investment, and offers a practical, flexible tool to guide industrial stakeholders in cost-effective battery energy storage system deployment.

Batteries↗

Multimodel ensemble predictions of precipitation using bayesian neural networks

Multimodel ensembling improves predictions and considers model uncertainties. In this study, we present a Bayesian Neural Network (BNN) ensemble approach for large-scale precipitation predictions based on a set of climate models. BNN infers spatiotemporally varying model weights and biases through the calibration against observations. This ensemble scheme of BNN sufficiently leverages individual model skill for accurate predictions as well as provides interpretability about which models contribute more to the ensemble prediction at which locations and times to inform model development. Additionally, BNN accurately quantifies epistemic uncertainty to avoid overconfident projections. We demonstrate BNN’s superior prediction performance to three state-of-the-art ensemble approaches and discuss its interpretability and uncertainty quantification.

Fan, Patrick↗

Multimodel Ensemble Predictions of Precipitation using Bayesian Neural Networks

Multimodel ensembling improves predictions and considers model uncertainties. In this study, we present a Bayesian Neural Network (BNN) ensemble approach for large-scale precipitation predictions based on a set of climate models. BNN infers spatiotemporally varying model weights and biases through the calibration against observations. This ensemble scheme of BNN sufficiently leverages individual model skill for accurate predictions as well as provides interpretability about which models contribute more to the ensemble prediction at which locations and times to inform model development. Additionally, BNN accurately quantifies epistemic uncertainty to avoid overconfident projections. We demonstrate BNN’s superior prediction performance to three state-of-the-art ensemble approaches and discuss its interpretability and uncertainty quantification.

Fan, Patrick↗

Stochastic Reconstruction of Thermal Protection Material Properties from Arc-Jet Experiments

Material response models are used to assess reliability using variances in the bond-line temperature predictions based on uncertainties in trajectory, aerothermal environment, and material properties. A key deficiency in the current approach is that input uncertainties are too often subjective, empirical, or ad-hoc, and are not rigorously linked to the arc-jet test data used to develop the TPS material model. While materials such as PICA are well understood, future missions may require more novel materials such as HEEET where unknown uncertainties have real consequences on the ability to assess reliability. A quantifiable estimate of reliability requires an iterative methodology where the parameters driving the variance in bond-line temperature (for example) are systematically identified. A test campaign to collect data or develop new models can then be identified to reduce those input uncertainties. A Bayesian inference loop defines these connections mathematically, i.e., prior knowledge about uncertainty is updated based on observation. While these concepts are well known (and often applied intuitively in a non-rigorous approach), only recent advances in reduced-order modelling have made them computationally viable methods for engineering. By replacing deterministic inverse methods with stochastic approaches, the hope is new materials proposed for future missions can more rapidly be developed with a greater understanding of the TPS material reliability. Two additional steps for the analysis of arc jet test data are discussed. The first is ability to construct a reduced-order model using material response simulations (Icarus/US3D) of the arc-jet test articles, and the second is the inclusion of this surrogate model in the Bayesian inversion process. Both capabilities will be demonstrated using prior PICA arc-jet test data. The quality of a surrogate model will be investigated and the variances on the calibrated material properties will be compared to our current understanding of the PICA material model.

Material response↗

Extending Parsimonious Bayesian Inference

Parsimonious Bayesian inference is a theoretical framework for efficient data assimilation that seeks to balance increased consistency between predictions and training data against corresponding increases in model complexity. Within this framework, over-training is understood as optimization that encodes excessive information within model parameters while only achieving small improvements between predictions and training data. This project aims to develop practical methods of limiting excess model information during optimization. One key observation is that practical heuristics for parsimonious learning in high-dimensions must balance expressivity, i.e. the ability of the model to capture diverse predictions with only a few non-zero parameters, against discoverability, i.e. the ability to train the model with gradient-based optimization and drive parameters to low information states. As such, we developed logical activation functions that are able to adaptively approximate arbitrary truth tables that define Boolean logic operations within a probabilistic framework. These functions have demonstrated the ability to learn exclusive disjunction (XOR) and conditioned disjunction (if [condition] then [result_if_true] else [result_if_false]) within a single layer of a neural network. To efficiently exploit these activation functions to drive parsimonious learning required several other advances within the domain of variational inference. The most efficient form of complexity suppression is structured sparsification, driving most model parameters to zero while achieving the structural coherence among nonzeros needed for bandwidth reduction. Such models are not only far more efficient at suppressing information-theoretic complexity, they also reduce the other forms of complexity (computations, communication, storage, and the number of dependencies needed to evaluate predictions). Aiming to support enhanced sparsification, this project examined new approaches to high-dimensional variational inference that allow us to calibrate and control parameter uncertainty during optimization. By identifying which parameters can sustain sparsifying perturbations with little impact on prediction quality, we can develop better pruning strategies by framing them as approximate Bayesian inference. These advances also open paths to mitigate concerns with deploying advanced learning methods in resource-constrained environments, such as running models on power-limited or communication-limited devices.

97 MATHEMATICS AND COMPUTING↗

Analytical gradient-based optimization of CALPHAD model parameters

The calibration of CALPHAD (CALculation of PHAse Diagrams) models involves the solution of a very challenging high-dimensional multiobjective optimization problem. Traditional approaches to parameter fitting predominantly rely on gradient-free methods, which while robust, are computationally inefficient and often scale poorly with model complexity. In this work, we introduce and demonstrate a generalizable framework for analytic gradient-based optimization of the parameters of the CALPHAD model enabled by the recently formalized Jansson derivative technique. This method allows for efficient evaluation of gradients of thermodynamic properties at equilibrium with respect to model parameters, even in the presence of arbitrarily complex internal degrees of freedom. Leveraging these semi-analytic gradients, we employ the conjugate gradient (CG) method to optimize thermodynamic model parameters for four binary alloy systems: Cu-Mg, Fe-Ni, Cr-Ni, and Cr-Fe. Across all systems, CG achieves comparable or superior optimality relative to Bayesian ensemble Markov Chain Monte Carlo (MCMC) with improvements in computational efficiency ranging from one to three orders of magnitude. Furthermore, our results establish a new paradigm for CALPHAD assessments in which high fidelity data-rich model calibration becomes tractable using deterministic gradient-informed algorithms.

CALPHAD↗

Fast strength model characterization using Bayesian statistics

A variety of flow stress models exist with new models constantly being developed. These models aim to approximate the strength of materials in a variety of regimes from quasistatic loading through shock scenarios. All models contain an array of parameters which need to be tuned to the material under study. Some models perform well under limited conditions, requiring adjustment of the parameters when venturing outside of those predefined ranges. Other models perform well over a wide range of conditions with a set of parameters, but may be outperformed by other models optimized on a tighter range of conditions. Recent research by Los Alamos demonstrated the ability to optimize the Johnson Cook (JC) model using a set of 3 plate-impact experiments on Aluminum. They utilized Bayesian statistics and emulation to determine optimal parameters for the model with a quantification of parameter uncertainty. We present an extension of this capability to incorporate velocimetry from plate-impact tests, stress-strain data from split Hopkinson pressure bar and quasistatic compression tests, plus profiles from Taylor cylinders in a unified fashion. Statistically robust comparisons of the performance and uncertainty of different realizations of the JC flow stress model were carried out based on calibration to several possible combinations of these three different experiment types.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Evidence cross-validation and Bayesian inference of MAST plasma equilibria

In this paper, current profiles for plasma discharges on the mega-ampere spherical tokamak are directly calculated from pickup coil, flux loop, and motional-Stark effect observations via methods based in the statistical theory of Bayesian analysis. By representing toroidal plasma current as a series of axisymmetric current beams with rectangular cross-section and inferring the current for each one of these beams, flux-surface geometry and q-profiles are subsequently calculated by elementary application of Biot-Savart's law. The use of this plasma model in the context of Bayesian analysis was pioneered by Svensson and Werner on the joint-European tokamak [Svensson and Werner,Plasma Phys. Controlled Fusion 50(8), 085002 (2008)]. In this framework, linear forward models are used to generate diagnostic predictions, and the probability distribution for the currents in the collection of plasma beams was subsequently calculated directly via application of Bayes' formula. In this work, we introduce a new diagnostic technique to identify and remove outlier observations associated with diagnostics falling out of calibration or suffering from an unidentified malfunction. These modifications enable a good agreement between Bayesian inference of the last-closed flux-surface with other corroborating data, such as that from force balance considerations using EFIT++[Appel et al., ''A unified approach to equilibrium reconstruction'' Proceedings of the 33rd EPS Conference on Plasma Physics (Rome, Italy, 2006)]. In addition, this analysis also yields errors on the plasma current profile and flux-surface geometry as well as directly predicting the Shafranov shift of the plasma core.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A full-spectral Bayesian reconstruction approach based on the material decomposition model applied in dual-energy computed tomography

Purpose: Dual-energy computed tomography (DECT) makes it possible to get two fractions of basis materials without segmentation. One is the soft-tissue equivalent water fraction and the other is the hard-matter equivalent bone fraction. Practical DECT measurements are usually obtained with polychromatic x-ray beams. Existing reconstruction approaches based on linear forward models without counting the beam polychromaticity fail to estimate the correct decomposition fractions and result in beam-hardening artifacts (BHA). The existing BHA correction approaches either need to refer to calibration measurements or suffer from the noise amplification caused by the negative-log preprocessing and the ill-conditioned water and bone separation problem. To overcome these problems, statistical DECT reconstruction approaches based on nonlinear forward models counting the beam polychromaticity show great potential for giving accurate fraction images.Methods: This work proposes a full-spectral Bayesian reconstruction approach which allows the reconstruction of high quality fraction images from ordinary polychromatic measurements. This approach is based on a Gaussian noise model with unknown variance assigned directly to the projections without taking negative-log. Referring to Bayesian inferences, the decomposition fractions and observation variance are estimated by using the joint maximum a posteriori (MAP) estimation method. Subject to an adaptive prior model assigned to the variance, the joint estimation problem is then simplified into a single estimation problem. It transforms the joint MAP estimation problem into a minimization problem with a nonquadratic cost function. To solve it, the use of a monotone conjugate gradient algorithm with suboptimal descent steps is proposed.Results: The performance of the proposed approach is analyzed with both simulated and experimental data. The results show that the proposed Bayesian approach is robust to noise and materials. It is also necessary to have the accurate spectrum information about the source-detector system. When dealing with experimental data, the spectrum can be predicted by a Monte Carlo simulator. For the materials between water and bone, less than 5% separation errors are observed on the estimated decomposition fractions.Conclusions: The proposed approach is a statistical reconstruction approach based on a nonlinear forward model counting the full beam polychromaticity and applied directly to the projections without taking negative-log. Compared to the approaches based on linear forward models and the BHA correction approaches, it has advantages in noise robustness and reconstruction accuracy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Uncertainty quantification of cinematic imaging for development of predictive simulations of turbulent combustion.

Recent advances in high frame rate complementary metal-oxide-semiconductor (CMOS) cameras coupled with high repetition rate lasers have enabled laser-based imaging measurements of the temporal evolution of turbulent reacting flows. This measurement capability provides new opportunities for understanding the dynamics of turbulence-chemistry interactions, which is necessary for developing predictive simulations of turbulent combustion. However, quantitative imaging measurements using high frame rate CMOS cameras require careful characterization of the their noise, non-linear response, and variations in this response from pixel to pixel. We develop a noise model and calibration tools to mitigate these problems and to enable quantitative use of CMOS cameras. We have demonstrated proof of principle for image de-noising using both wavelet methods and Bayesian inference. The results offer new approaches for quantitative interpretation of imaging measurements from noisy data acquired with non-linear detectors. These approaches are potentially useful in many areas of scientific research that rely on quantitative imaging measurements.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Probabilistic methods for sensitivity analysis and calibration in the NASA challenge problem

In this study, a series of algorithms are proposed to address the problems in the NASA Langley Research Center Multidisciplinary Uncertainty Quantification Challenge. A Bayesian approach is employed to characterize and calibrate the epistemic parameters based on the available data, whereas a variance-based global sensitivity analysis is used to rank the epistemic and aleatory model parameters. A nested sampling of the aleatory–epistemic space is proposed to propagate uncertainties from model parameters to output quantities of interest.

97 MATHEMATICS AND COMPUTING↗