Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “bayesian”

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 1,027 records · Page 57

Optimizing collimator positions using bayesian optimization in the Fermilab MI-8 transfer line

Collimators are used to minimize losses and to remove particles that would otherwise get lost downstream and irradiate the machine. Finding the optimal jaw positions is time consuming and with the upstream beam properties changing, the collimation settings would need to be readjusted each time. Therefore, a method to optimize collimator positions and to operate them at full capacity in a short time is required for loss control downstream. A study of collimator positions was conducted and a machine learning (ML) model was developed to predict optimal collimator positions. Bayesian Optimization (BO) was used to calculate new jaw positions from the ML model. The results of BO and usage of ML for better performance of the collimation system are presented in this paper.

Babacan, Betiay [Fermilab]↗

Fermilab Booster loss modelling and rebalancing using Bayesian methods

To meet PIP-II upgrade requirements, Fermilab Booster losses need to be reduced by 50% compared to present levels. So far, simulations are not good enough to predict loss patterns. Thus, an extensive Booster tune up will be necessary to achieve required performance. In this paper we present an effort to build a data-driven loss model using Bayesian techniques, and subsequently to rebalance losses for higher trip margins. We first created several sets of spatially and temporally isolated orbit and optics knobs, and trained Gaussian process models for each beam loss monitor as well as beam current. Novel techniques of uncertainty constraints and approximate GP fitting were introduced to handle safety and timing requirements. We then performed single and multi-objective tuning using scalarized objectives comprised of critical beam loss locations. We achieved significant rebalancing of losses, increasing margins by 25%, as well as an overall improvement in transmission efficiency of 0.4%. Automated data collection is being developed so that more accurate surrogate models can be trained over time.

Kuklev, N. [Fermilab]↗

Three-dimensional multiphase segmentation of X-ray CT data of porous materials using a Bayesian Markov random field framework

Advancements in noninvasive imaging methods such as X-ray computed tomography (CT) have led to a recent surge of applications in porous media research with objectives ranging from theoretical aspects of pore-scale fluid and interfacial dynamics to practical applications such as enhanced oil recovery and advanced contaminant remediation. While substantial efforts and resources have been devoted to advance CT technology, microscale analysis, and fluid dynamics simulations, the development of efficient and stable three-dimensional multiphase image segmentation methods applicable to large data sets is lacking. To eliminate the need for wet-dry or dual-energy scans, image alignment, and subtraction analysis, commonly applied in X-ray micro-CT, a segmentation method based on a Bayesian Markov random field (MRF) framework amenable to true three-dimensional multiphase processing was developed and evaluated. Furthermore, several heuristic and deterministic combinatorial optimization schemes required to solve the labeling problem of the MRF image model were implemented and tested for computational efficiency and their impact on segmentation results. Test results for three grayscale data sets consisting of dry glass beads, partially saturated glass beads, and partially saturated crushed tuff obtained with synchrotron X-ray micro-CT demonstrate great potential of the MRF image model for three-dimensional multiphase segmentation. While our results are promising and the developed algorithm is stable and computationally more efficient than other commonly applied porous media segmentation models, further potential improvements exist for fully automated operation.

42 ENGINEERING↗

Exploring unsupervised top tagging using Bayesian inference

Recognizing hadronically decaying top-quark jets in a sample of jets, or even its total fraction in the sample, is an important step in many LHC searches for Standard Model and Beyond Standard Model physics as well. Although there exists outstanding top-tagger algorithms, their construction and their expected performance rely on Montecarlo simulations, which may induce potential biases. For these reasons we develop two simple unsupervised top-tagger algorithms based on performing Bayesian inference on a mixture model. In one of them we use as the observed variable a new geometrically-based observable \tilde{A}_{3} A ̃ 3 , and in the other we consider the more traditional \tau_{3}/\tau_{2} τ 3 / τ 2 N N -subjettiness ratio, which yields a better performance. As expected, we find that the unsupervised tagger performance is below existing supervised taggers, reaching expected Area Under Curve AUC \sim 0.80-0.81 ∼ 0.80 − 0.81 and accuracies of about 69% - − 75% in a full range of sample purity. However, these performances are more robust to possible biases in the Montecarlo that their supervised counterparts. Our findings are a step towards exploring and considering simpler and unbiased taggers.

Alvarez, Ezequiel↗

Large-Scale Optimization for Bayesian Inference in Complex Systems

The SAGUARO (Scalable Algorithms for Groundwater Uncertainty Analysis and Robust Optimization) Project focused on the development of scalable numerical algorithms for large-scale Bayesian inversion in complex systems that capitalize on advances in large-scale simulation-based optimization and inversion methods. The project was a collaborative effort among MIT, the University of Texas at Austin, Georgia Institute of Technology, and Sandia National Laboratories. The research was directed in three complementary areas: efficient approximations of the Hessian operator, reductions in complexity of forward simulations via stochastic spectral approximations and model reduction, and employing large-scale optimization concepts to accelerate sampling. The MIT₋Sandia component of the SAGUARO Project addressed the intractability of conventional sampling methods for large-scale statistical inverse problems by devising reduced-order models that are faithful to the full-order model over a wide range of parameter values; sampling then employs the reduced model rather than the full model, resulting in very large computational savings. Results indicate little effect on the computed posterior distribution. On the other hand, in the Texas₋Georgia Tech component of the project, we retain the full-order model, but exploit inverse problem structure (adjoint-based gradients and partial Hessian information of the parameter-to-observation map) to implicitly extract lower dimensional information on the posterior distribution; this greatly speeds up sampling methods, so that fewer sampling points are needed. We can think of these two approaches as "reduce then sample" and "sample then reduce." In fact, these two approaches are complementary, and can be used in conjunction with each other. Moreover, they both exploit deterministic inverse problem structure, in the form of adjoint-based gradient and Hessian information of the underlying parameter-to-observation map, to achieve their speedups.

97 MATHEMATICS AND COMPUTING↗

Bayesian calibration of the Community Land Model using surrogates

We present results from the Bayesian calibration of hydrological parameters of the Community Land Model (CLM), which is often used in climate simulations and Earth system models. A statistical inverse problem is formulated for three hydrological parameters, conditional on observations of latent heat surface fluxes over 48 months. Our calibration method uses polynomial and Gaussian process surrogates of the CLM, and solves the parameter estimation problem using a Markov chain Monte Carlo sampler. Posterior probability densities for the parameters are developed for two sites with different soil and vegetation covers. Our method also allows us to examine the structural error in CLM under two error models. We find that surrogate models can be created for CLM in most cases. The posterior distributions are more predictive than the default parameter values in CLM. Climatologically averaging the observations does not modify the parameters' distributions significantly. The structural error model reveals a correlation time-scale which can be used to identify the physical process that could be contributing to it. While the calibrated CLM has a higher predictive skill, the calibration is under-dispersive.

54 ENVIRONMENTAL SCIENCES↗

A Bayesian inversion framework for yield and height-of-burst/depth-of-burial for near-surface explosions

A Bayesian inversion framework is presented to estimate the yield of an explosion and height-of-burst/depth-of-burial (HOB/DOB) using seismic and air pressure data. This is accomplished by first calibrating the parameters in the forward models that relate the observations to the yield and HOB/DOB and then using the calibrated model to estimate yield and HOB/DOB associated with a new set of seismic and air pressure observations. The MCMC algorithms required to perform these steps are outlined, and the results with real data are shown. Finally, an extension is proposed for a case when clustering in the seismic displacement occurs as a function of different types of rock and other factors.

58 GEOSCIENCES↗

Bayesian Regression of Thermodynamic Models of Redox Active Materials

Finding a suitable functional redox material is a critical challenge to achieving scalable, economically viable technologies for storing concentrated solar energy in the form of a defected oxide. Demonstrating e ectiveness for thermal storage or solar fuel is largely accomplished by using a thermodynamic model derived from experimental data. The purpose of this project is to test the accuracy of our regression model on representative data sets. Determining the accuracy of the model includes parameter tting the model to the data, comparing the model using di erent numbers of param- eters, and analyzing the entropy and enthalpy calculated from the model. Three data sets were considered in this project: two demonstrating materials for solar fuels by wa- ter splitting and the other of a material for thermal storage. Using Bayesian Inference and Markov Chain Monte Carlo (MCMC), parameter estimation was preformed on the three data sets. Good results were achieved, except some there was some deviations on the edges of the data input ranges. The evidence values were then calculated in a variety of ways and used to compare models with di erent number of parameters. It was believed that at least one of the parameters was unnecessary and comparing evidence values demonstrated that the parameter was need on one data set and not signi cantly helpful on another. The entropy was calculated by taking the derivative in one variable and integrating over another. and its uncertainty was also calculated by evaluating the entropy over multiple MCMC samples. Afterwards, all the parts were written up as a tutorial for the Uncertainty Quanti cation Toolkit (UQTk).

36 MATERIALS SCIENCE↗

Exploiting Cross-sensitivity by Bayesian Decoding of Mixed Potential Sensor Arrays

LANL mixed-potential electrochemical sensor (MPES) device arrays were coupled with advanced Bayesian inference treatment of the physical model of relevant sensor-analyte interactions. We demonstrated that our approach could be used to uniquely discriminate the composition of ternary gas sensors with three discreet MPES sensors with an average error of less than 2%. We also observed that the MPES exhibited excellent stability over a year of operation at elevated temperatures in the presence of test gases.

47 OTHER INSTRUMENTATION↗

Application of Bayesian Model Selection for Metal Yield Models using ALEGRA and Dakota

This report introduces the concepts of Bayesian model selection, which provides a systematic means of calibrating and selecting an optimal model to represent a phenomenon. This has many potential applications, including for comparing constitutive models. The ideas described herein are applied to a model selection problem between different yield models for hardened steel under extreme loading conditions.

97 MATHEMATICS AND COMPUTING↗

Bayesian characterization and forward propagation of the uncertainty in thermodynamic models for redox active materials

Demonstrating the thermodynamic efficiency of hydrogen conversion processes using various materials is a critical step in developing new technologies for storing concentrated solar energy, and is largely accomplished by using a thermodynamic model derived from experimental data. A main goal of this project is to calculate the uncertainty of the thermodynamic efficiency by calculating the uncertainty of the components that feed into the efficiency. Many different models and data sets were used to test the workflow. First, the models were fit to the data using a Bayesian Inference and a method called Markov Chain Monte Carlo (MCMC), which found the maximum a priori parameters, and a posterior probability distribution of the parameters. Next, the different models were compared to each other using model evidence values. It was found that for cleaner data sets, overfitting had not yet been reached, and the most complicated model was ideal, but on the noisier data sets, the less complex models were favored because the more complicated models resulted in overfitting. Next, forward propagation was used to calculate the enthalpy change and its associated uncertainty. A few variations on the models were tried, such as fitting in a different variable, producing negligible or negative effects on the fits of the models. Thus, the original models were used. A sensitivity analysis was performed, and used to calculate the model error. On the cleaner data sets, there was very minimal experimental noise, and thus, all resulting error was from the model. With consideration of the model error, the models fit the data very well, and the simpler model had a high model error, as expected. All these components will then be used to calculate the thermodynamic efficiency of the different materials.

14 SOLAR ENERGY↗

Joint Bayesian Inference for Near-surface Explosion Yield

Geophysical observations of an explosion scale with the yield of the explosion. Guided by the principal of yield-scaling and observed relationships in the data, we defined and parameterized seismoacoustic forward models in Ford et al. [2014]. This work extends the model with additional data features of maximum velocity and positive phase duration for the seismic and acoustic data types, respectively, and a new optic data type. Parameters for the seismoacoustic models are estimated via Bayesian inference. The parameterized models can then be used to infer the yield of a new explosion.

58 GEOSCIENCES↗

Bayesian Optimal Sensor Augmentation Via Estimated Mutual Information

We consider the problem of designing a sensor network to most efficiently locate the source of a seismic event. The left panel of Figure 1 shows 982 possible sensor locations in a region of interest. The objective is to infer the unknown location of an earthquake, indicated by the gold dot for illustration purposes. Assume that three sensors have already been placed for data collection. Travel times of the P-wave from the source to the three sensors are observed and used to infer the unknown source location via a Bayesian statistical model described subsequently. The right panel of Figure 1 shows samples from the resulting posterior distribution for source location along with a posterior estimate of this location.

47 OTHER INSTRUMENTATION↗

A Bayesian Neural Network Ensemble Approach for Improving Large-Scale Streamflow Predictability

Predictive modeling through the use of AI techniques and AI-derived model components; the use of AI and other tools to design a prediction system comprising of a hierarchy of models (e.g.,AI driven model/component/parameterization selection). We propose a Bayesian neural network ensemble approach for improving large-scale streamflow predictability and understanding in a changing environment that combines multiple Earth system land model predictions by calculating spatiotemporally varying model weights and biases while accounting for various types of observations at multiple scales with uncertainty.

58 GEOSCIENCES↗

Using Bayesian Methodology to Estimate Liquefied Natural Gas Leak Frequencies

This analysis provides estimates on the leak frequencies of nine components found in liquefied natural gas (LNG) facilities. Data was taken from a variety of sources, with 25 different data sets included in the analysis. A hierarchical Bayesian model was used that assumes that the log leak frequency follows a normal distribution and the logarithm of the mean of this normal distribution is a linear function of the logarithm of the fractional leak area. This type of model uses uninformed prior distributions that are updated with applicable data. Separate models are fit for each component listed. Five order-of-magnitude fractional leak areas are considered, based on the flow area of the component. Three types of supporting analyses were performed: sensitivity of the model to the data set used, sensitivity of the leak frequency estimates to differences in the model structure or prior distributions, and sufficiency of sample sized used for convergence. Recommended leak frequency distributions for all component types and leak sizes are given. These leak frequency predictions can be used for quantitative risk assessments in the future.

03 NATURAL GAS↗

Semi-supervised Bayesian Low-shot Learning

Deep neural networks (NNs) typically outperform traditional machine learning (ML) approaches for complicated, non-linear tasks. It is expected that deep learning (DL) should offer superior performance for the important non-proliferation task of predicting explosive device configuration based upon observed optical signature, a task which human experts struggle with. However, supervised machine learning is difficult to apply in this mission space because most recorded signatures are not associated with the corresponding device description, or “truth labels.” This is challenging for NNs, which traditionally require many samples for strong performance. Semi-supervised learning (SSL), low-shot learning (LSL), and uncertainty quantification (UQ) for NNs are emerging approaches that could bridge the mission gaps of few labels and rare samples of importance. NN explainability techniques are important in gaining insight into the inferential feature importance of such a complex model. In this work, SSL, LSL, and UQ are merged into a single framework, a significant technical hurdle not previously demonstrated. Exponential Average Adversarial Training (EAAT) and Pairwise Neural Networks (PNNs) are chosen as the SSL and LSL methods of choice. Permutation feature importance (PFI) for functional data is used to provide explainability via the Variable importance Explainable Elastic Shape Analysis (VEESA) pipeline. A variety of uncertainty quantification approaches are explored: Bayesian Neural Networks (BNNs), ensemble methods, concrete dropout, and evidential deep learning. Two final approaches, one utilizing ensemble methods and one utilizing evidential learning, are constructed and compared using a well-quantified synthetic 2D dataset along with the DIRSIG Megascene.

97 MATHEMATICS AND COMPUTING↗

A consistent, Bayesian, approach to the cross section probability distribution in the unresolved resonance region

The cross sections of neutron-induced reactions can be divided into three energy ranges: the resolved resonance region (RRR), the unresolved resonance region (URR), and the fast region. In general, the cross sections in the URR show significant fluctuations that cannot be predicted and cannot be experimentally resolved, thus, it is commonly assumed that the cross section at a specific energy is given by a probability distribution function (PDF) over a range of values that can span several orders of magnitude. The current methodology used to describe such behavior is to construct the PDF by stochastically generating resonance ladders and numerically measuring the PDF. The resonance ladders are sampled using known resonance statistical properties and average resonance widths and spacings extrapolated from the RRR. Although this is a standard and widely used technique, it is computationally very expensive, therefore, an alternative, analytical, approach would be preferable due to the considerable speed up of the computational time in real life applications. Moreover, the current methodology does not take into account existing experimental data, such for total and capture cross sections, that are available for many nuclei. Finally, this approach was developed to be used in reactor-scale applications and it is not suited for use in single-event applications. In this work we will rethink the entire approach to the PDF construction using a Bayesian mindset. This will allow us to provide a different definition of the PDF that allows a much faster calculation of the higher-temperature PDFs and a proper combination of theoretical and experimental PDFs following the probability theory. We will also show that our definition is well suited for single-event applications and we will make an explicit connection between our method and the standard approach. We do this by showing that the central limit theorem applies and our method leads to the same PDF obtained with the standard methodology, for a large number of events per history.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗