Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “uncertainty Quantification”

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 91 records · Page 5

Adaptive polynomial chaos techniques for uncertainty quantification of a gas cooled fast reactor transient

Uncertainty quantification plays an increasingly important role in the nuclear community, especially with the rise of Best Estimate Plus Uncertainty methodologies. Sensitivity analysis, surrogate models, Monte Carlo sampling and several other techniques can be used to propagate input uncertainties. In recent years however polynomial chaos expansion has become a popular alternative providing high accuracy at affordable computational cost. This paper presents such polynomial chaos (PC) methods using adaptive sparse grids and adaptive basis set construction, together with an application to a Gas Cooled Fast Reactor transient. Comparison is made between a new sparse grid algorithm and the traditionally used technique proposed by Gerstner. An adaptive basis construction method is also introduced and is proved to be advantageous both from an accuracy and a computational point of view. As a demonstration the uncertainty quantification of a 50% loss of flow transient in the GFR2400 Gas Cooled Fast Reactor design was performed using the CATHARE code system. The results are compared to direct Monte Carlo sampling and show the superior convergence and high accuracy of the polynomial chaos expansion. Since PC techniques are easy to implement, they can offer an attractive alternative to traditional techniques for the uncertainty quantification of large scale problems. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Interpretable Uncertainty Quantification in AI for HEP

Estimating uncertainty is at the core of performing scientific measurements in HEP: a measurement is not useful without an estimate of its uncertainty. The goal of uncertainty quantification (UQ) is inextricably linked to the question, "how do we physically and statistically interpret these uncertainties?" The answer to this question depends not only on the computational task we aim to undertake, but also on the methods we use for that task. For artificial intelligence (AI) applications in HEP, there are several areas where interpretable methods for UQ are essential, including inference, simulation, and control/decision-making. There exist some methods for each of these areas, but they have not yet been demonstrated to be as trustworthy as more traditional approaches currently employed in physics (e.g., non-AI frequentist and Bayesian methods). Shedding light on the questions above requires additional understanding of the interplay of AI systems and uncertainty quantification. We briefly discuss the existing methods in each area and relate them to tasks across HEP. We then discuss recommendations for avenues to pursue to develop the necessary techniques for reliable widespread usage of AI with UQ over the next decade.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

CALPHAD Uncertainty Quantification and TDBX

CALPHAD uncertainty quantification (UQ) is the foundation of materials design with quantified confidence. We report a framework and software packages to enable CALPHAD UQ assessment and calculation using commercial CALPHAD software (Thermo-Calc). This Bayesian inference framework is coupled with a Markov chain Monte Carlo algorithm to establish uncertainty traces with a given thermodynamic database file (TDB) and corresponding experimental data points. This general framework is demonstrated with the Ni–Cr binary system. The algorithm is firstly validated on synthetic data with known ground truth. Then it is applied to real experimental data to generate posterior traces. We develop a file format named TDBX, which provides a single source of truth by combining the original TDB content and the traces for each assessed Gibbs energy parameter. CALPHAD UQ calculations are performed based on the TDBX file, from which uncertainties for phase boundaries, enthalpy curves, and solidification range are collected as examples of basic design parameters. This TDBX file with corresponding scripts are made open-source. Finally, the combination of CALPHAD UQ assessments and calculations connected by TDBX supports uncertainty-assisted modeling, enabling the integrated application of modern design with uncertainty methodologies to computational materials design.

36 MATERIALS SCIENCE↗

Adaptive Sparse-Grid Stochastic Collocation Uncertainty Quantification Convergence for Multigroup Diffusion

Advanced methods in uncertainty quantification for numerical models in computational physics continue to gain acceptance in nuclear modeling. Previously, the efficiency of sparse-grid stochastic collocation in comparison with Monte Carlo for uncertainty quantification through convergence studies was demonstrated. An adaptive method for anisotropic sparse grid collocation and its convergence compared to previous efforts is considered. The physical system in consideration is a two-dimensional quarter-core reactor, consisting of 5 materials distributed in 121 regions. The two-group neutron diffusion criticality approximation is solved. Vacuum boundaries are applied on the top and right, and reflecting boundaries on the bottom and left. The material cross sections, neutron multiplication factors, and diffusion coefficients are potential sources of uncertainty. Uniformly-distributed uncertainty within 5% of the reference values is introduced for material properties. The distribution of the uncertain parameters make up the uncertainty space Γ included in R{sup N}, where N is the number of uncertain parameters. The input parameter uncertainties are independently distributed. As expected, the convergence of Monte Carlo on the benchmark as a function of computational solves is nearly linear and lethargic compared to the other methods. The static isotropic index set, in which each input dimension is treated with equal importance, converges much more efficiently than Monte Carlo, but lacks the benefit of emphasizing some dimensions over others. The anisotropic case is most efficient, where first-order sensitivity information was used to determine anisotropic weights for polynomials to use in the polynomial chaos expansion. Interestingly, the adaptive algorithm converges slightly less efficiently than the anisotropic case, while still performing better than the isotropic case. Because the adaptive case has to determine the importance of each input dimension, it occasionally predicts inaccurately and wastes computation solves compared to the more ideal anisotropic case. However, in circumstances when an ideal anisotropy is not understood a priori, the adaptive algorithm may be optimally efficient in representing the original model. Given the potential loss of efficiency when using a poorly-chosen anisotropic weighting, the adaptive algorithm may be an effective choice for initial uncertainty quantification in problems with a low-dimensionality input space. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

End-To-End Uncertainty Quantification with Analytical Derivatives for Design Under Uncertainty

Uncertainty quantification (UQ) is a rapidly growing and evolving discipline, especially within the aerospace community. Performing analysis with UQ can provide decision makers with a wealth of information about a candidate design. However, the value of UQ is fully realized when the information gained during UQ analysis is leveraged in a feedback loop of a design optimization process, often referred to as design under uncertainty. Although design under uncertainty can be a powerful risk mitigation technique, there are a number of roadblocks that prevent its implementation. Two primary factors are computational costs and added complexity of the analysis. High fidelity simulations on the order tens of uncertain variables quickly become computationally infeasible. Also, implementing UQ into an existing multidisciplinary design and optimization (MDO) process often requires extensive knowledge of the UQ methods and careful treatment of the problem formulation. The objective of this work is to address these two primary roadblocks and enable practitioners to efficiently perform design under uncertainty with limited knowledge of the UQ discipline. Methods outlined in this paper demonstrate MDO incorporating UQ into the design process, leveraging an analytic derivative tool chain through the entire optimization. The proposed approach leverages machine learning techniques to generate a differentiable confidence interval output from polynomial chaos models. This technique, coupled with the incorporation of analytical derivatives through the Polynomial Chaos Expansion (PCE) process, eliminates the need to estimate derivatives which are usually obtained from finite difference, complex step, or similar methods. Developing a differentiable confidence interval allows mixed uncertainty problems (both epistemic and aleatory) to be modeled. Without such modeling, these problems cannot accurately predict objective functions containing statistical quantities such as mean and variance. The addition of analytic derivatives to a polynomial chaos-based UQ method decreases the computational costs of performing design under uncertainty by orders of magnitude in comparison with methods such as complex step. The method and codes developed are modular in nature and are a drop-in solution for design under uncertainty within existing MDO problems. A low-fidelity analytical multidisciplinary optimization under uncertainty for a wing design in OpenMDAO is detailed in this paper. This demonstration case will include both objective functions and constraints which are influenced by uncertain parameters.

Ben D Phillips↗

End-To-End Uncertainty Quantification with Analytical Derivatives for Design Under Uncertainty

Uncertainty quantification (UQ) is a rapidly growing and evolving discipline, especially within the aerospace community. Performing analysis with UQ can provide decision makers with a wealth of information about a candidate design. However, the value of UQ is fully realized when the information gained during UQ analysis is leveraged in a feedback loop of a design optimization process, often referred to as design under uncertainty. Although design under uncertainty can be a powerful risk mitigation technique, there are a number of roadblocks that prevent its implementation. Two primary factors are computational costs and added complexity of the analysis. High fidelity simulations on the order tens of uncertain variables quickly become computationally infeasible. Also, implementing UQ into an existing multidisciplinary design and optimization (MDO) process often requires extensive knowledge of the UQ methods and careful treatment of the problem formulation. The objective of this work is to address these two primary roadblocks and enable practitioners to efficiently perform design under uncertainty with limited knowledge of the UQ discipline. Methods outlined in this paper demonstrate MDO incorporating UQ into the design process, leveraging an analytic derivative tool chain through the entire optimization. The proposed approach leverages machine learning techniques to generate a differentiable confidence interval output from polynomial chaos models. This technique, coupled with the incorporation of analytical derivatives through the Polynomial Chaos Expansion (PCE) process, eliminates the need to estimate derivatives which are usually obtained from finite difference, complex step, or similar methods. Developing a differentiable confidence interval allows mixed uncertainty problems (both epistemic and aleatory) to be modeled. Without such modeling, these problems cannot accurately predict objective functions containing statistical quantities such as mean and variance. The addition of analytic derivatives to a polynomial chaos-based UQ method decreases the computational costs of performing design under uncertainty by orders of magnitude in comparison with methods such as complex step. The method and codes developed are modular in nature and are a drop-in solution for design under uncertainty within existing MDO problems. A low-fidelity analytical multidisciplinary optimization under uncertainty for a wing design in OpenMDAO is detailed in this paper. This demonstration case will include both objective functions and constraints which are influenced by uncertain parameters.

Ben Phillips↗

Deep neural network uncertainty quantification for LArTPC reconstruction

We evaluate uncertainty quantification (UQ) methods for deep learning applied to liquid argon time projection chamber (LArTPC) physics analysis tasks. As deep learning applications enter widespread usage among physics data analysis, neural networks with reliable estimates of prediction uncertainty and robust performance against overconfidence and out-of-distribution (OOD) samples are critical for their full deployment in analyzing experimental data. While numerous UQ methods have been tested on simple datasets, performance evaluations for more complex tasks and datasets are scarce. Here we assess the application of selected deep learning UQ methods on the task of particle classification using the PiLArNet monte carlo 3D LArTPC point cloud dataset. We observe that UQ methods not only allow for better rejection of prediction mistakes and OOD detection, but also generally achieve higher overall accuracy across different task settings. We assess the precision of uncertainty quantification using different evaluation metrics, such as distributional separation of prediction entropy across correctly and incorrectly identified samples, receiver operating characteristic curves (ROCs), and expected calibration error from observed empirical accuracy. We conclude that ensembling methods can obtain well calibrated classification probabilities and generally perform better than other existing methods in deep learning UQ literature.

47 OTHER INSTRUMENTATION↗

Bi-fidelity variational auto-encoder for uncertainty quantification

Quantifying the uncertainty of quantities of interest (QoIs) from physical systems is a primary objective in model validation. However, achieving this goal entails balancing the need for computational efficiency with the requirement for numerical accuracy. To address this trade-off, we propose a novel bi-fidelity formulation of variational auto-encoders (BF-VAE) designed to estimate the uncertainty associated with a QoI from low-fidelity (LF) and high-fidelity (HF) samples of the QoI. Here, this model allows for the approximation of the statistics of the HF QoI by leveraging information derived from its LF counterpart. Specifically, we design a bi-fidelity auto-regressive model in the latent space which is integrated within the VAE’s probabilistic encoder–decoder structure. An effective algorithm is proposed to maximize the variational lower bound of the HF log-likelihood in the presence of limited HF data, resulting in the synthesis of HF realizations with a reduced computational cost. Additionally, we introduce the concept of the bi-fidelity information bottleneck (BF-IB) to provide an information-theoretic interpretation of the proposed BF-VAE model. Our numerical results demonstrate that the BF-VAE leads to considerably improved accuracy, as compared to a VAE trained using only HF data, when limited HF data is available.

42 ENGINEERING↗

Uncertainty quantification and sensitivity analysis applications to fuel performance modeling

Best-estimate fuel performance codes such as BISON currently under development at the Idaho National Laboratory, utilize empirical and mechanistic lower-length-scale informed correlations to predict fuel behavior under normal operating and accident reactor conditions. Traditionally, best-estimate results are presented using the correlations with no quantification of the uncertainty in the output metrics of interest. However, there are associated uncertainties in the input parameters and correlations used to determine the behavior of the fuel and cladding under irradiation. Therefore, it is important to perform uncertainty quantification and include confidence bounds on the output metrics that take into account the uncertainties in the inputs. In addition, sensitivity analyses can be performed to determine which input parameters have the greatest influence on the outputs. In this paper we couple the BISON fuel performance code to the DAKOTA uncertainty analysis software to analyze a representative fuel performance problem. The case studied in this paper is based upon rod 1 from the IFA-432 integral experiment performed at the Halden Reactor in Norway. The rodlet is representative of a BWR fuel rod. The input parameters uncertainties are broken into three separate categories including boundary condition uncertainties (e.g., power, coolant flow rate), manufacturing uncertainties (e.g., pellet diameter, cladding thickness), and model uncertainties (e.g., fuel thermal conductivity, fuel swelling). Utilizing DAKOTA, a variety of statistical analysis techniques are applied to quantify the uncertainty and sensitivity of the output metrics of interest. Specifically, we demonstrate the use of sampling methods, polynomial chaos expansions, surrogate models, and variance-based decomposition. The output metrics investigated in this study are the fuel centerline temperature, cladding surface temperature, fission gas released, and fuel rod diameter. The results highlighted the importance of the thermal conductivity of the fuel, the thermal conductivity of the cladding, the gap thickness, the fuel thermal expansion, and the system power factor as influential parameters.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

HPC Analytics Support. Requirements for Uncertainty Quantification Benchmarks

This report outlines techniques for extending benchmark generation products so they support uncertainty quantification by benchmarked systems. We describe how uncertainty quantification requirements can be presented to candidate analytical tools supporting SPARQL. We describe benchmark data sets for evaluating uncertainty quantification, as well as an approach for using our benchmark generator to produce data sets for generating benchmark data sets.

benchmark↗

Uncertainty quantification and sensitivity analysis with CASL Core Simulator VERA-CS

Uncertainty quantification and sensitivity analysis are important for nuclear reactor safety design and analysis. A 2x2 fuel assembly core design was developed and simulated by the Virtual Environment for Reactor Applications, Core Simulator (VERA-CS) coupled neutronics and thermal-hydraulics code under development by the Consortium for Advanced Simulation of Light Water Reactors (CASL). An approach to uncertainty quantification and sensitivity analysis with VERA-CS was developed and a new toolkit was created to perform uncertainty quantification and sensitivity analysis with fourteen uncertain input parameters. Furthermore, the minimum departure from nucleate boiling ratio (MDNBR), maximum fuel center-line temperature, and maximum outer clad surface temperature were chosen as the selected figures of merit. Pearson, Spearman, and partial correlation coefficients were considered for all of the figures of merit in sensitivity analysis and coolant inlet temperature was consistently the most influential parameter. We used parameters as inputs to the critical heat flux calculation with the W-3 correlation were shown to be the most influential on the MDNBR, maximum fuel center-line temperature, and maximum outer clad surface temperature.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Uncertainty Quantification for Monte Carlo simulations - Paper 18

Uncertainty quantification (UQ) generally speaking refers to a wide variety of specific problems, also pertaining different scientific areas: then studies on Uncertainty Quantification involve many different methodologies, techniques and even terminologies: to be specific we are interested in studying how uncertainties on input physical data needed for Monte Carlo simulations transfer into uncertainties on the physical output of the simulation. We mention that other sources of uncertainty are be present in a simulation environment: they can be ascribed either to model inadequacy (for instance when one neglects atomic lattice at very low energies) or to the numerical procedure involved: for the case of MC simulation the latter corresponds to the statistical error affecting the simulation itself. So in general parameter uncertainties are mixed with other kind of uncertainties. The problem then remains seemingly quite complicated, even in this more specific domain: moreover we should better specify the meaning of the rather undetermined words physical data needed for Monte Carlo simulation. In fact they may refer both those data needed to specify the experimental setup that is going to be simulated - like, e.g., geometrical or composition data, but including also some external conditions applied to the system, like temperature, pressure, electromagnetic fields and so on - and physical data - cross sections or often physical models - required to describe particle transport. All these data are generally affected by uncertainties, which through the process of simulation transfer into uncertainty on its physical outcome. Then, from the MC user's own perspective, input data refers both to those that are under his direct experimental control and to many others, whose values and uncertainties will affect the result of the simulation. In this contribution we will summarize previous results, together with some new ones. (authors)

61 RADIATION PROTECTION AND DOSIMETRY↗

Pre-conditioned BFGS-based uncertainty quantification in elastic full-waveform inversion

SUMMARY Full-waveform inversion has become an essential technique for mapping geophysical subsurface structures. However, proper uncertainty quantification is often lacking in current applications. In theory, uncertainty quantification is related to the inverse Hessian (or the posterior covariance matrix). Even for common geophysical inverse problems its calculation is beyond the computational and storage capacities of the largest high-performance computing systems. In this study, we amend the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm to perform uncertainty quantification for large-scale applications. For seismic inverse problems, the limited-memory BFGS (L-BFGS) method prevails as the most efficient quasi-Newton method. We aim to augment it further to obtain an approximate inverse Hessian for uncertainty quantification in FWI. To facilitate retrieval of the inverse Hessian, we combine BFGS (essentially a full-history L-BFGS) with randomized singular value decomposition to determine a low-rank approximation of the inverse Hessian. Setting the rank number equal to the number of iterations makes this solution efficient and memory-affordable even for large-scale problems. Furthermore, based on the Gauss–Newton method, we formulate different initial, diagonal Hessian matrices as pre-conditioners for the inverse scheme and compare their performances in elastic FWI applications. We highlight our approach with the elastic Marmousi benchmark model, demonstrating the applicability of pre-conditioned BFGS for large-scale FWI and uncertainty quantification.

58 GEOSCIENCES↗

Dimensionality reduction for uncertainty quantification of nuclear engineering models.

The task of uncertainty quantification consists of relating the available information on uncertainties in the model setup to the resulting variation in the outputs of the model. Uncertainty quantification plays an important role in complex simulation models of nuclear engineering, where better understanding of uncertainty results in greater confidence in the model and in the improved safety and efficiency of engineering projects. In our previous work, we have shown that the effect of uncertainty can be approximated by polynomial regression with derivatives (PRD): a hybrid regression method that uses first-order derivatives of the model output as additional fitting conditions for a polynomial expansion. Numerical experiments have demonstrated the advantage of this approach over classical methods of uncertainty analysis: in precision, computational efficiency, or both. To obtain derivatives, we used automatic differentiation (AD) on the simulation code; hand-coded derivatives are acceptable for simpler models. We now present improvements on the method. We use a tuned version of the method of snapshots, a technique based on proper orthogonal decomposition (POD), to set up the reduced order representation of essential information on uncertainty in the model inputs. The automatically obtained sensitivity information is required to set up the method. Dimensionality reduction in combination with PRD allows analysis on a larger dimension of the uncertainty space (>100), at modest computational cost.

42 ENGINEERING↗

Uncertainty Quantification for Neutron Shield Using Convolutional Neural Networks

Uncertainty quantification from radiation transport calculations was conducted using a Bayesian inference approach. A surrogate model, using a convolutional neural network, was employed to emulate the neutron fluence, which was simulated with a Monte Carlo radiation transport model. This allowed for a computationally cheap approach to evaluate input parameters and to sample their corresponding posterior probability distributions. Experimental data from the literature were employed to perform uncertainty quantification studies for concrete shields. As a result, the method is a nonintrusive approach that enables studies with multiple input parameters and can be applied to any radiation transport model.

Bayesian inference↗

Uncertainty Quantification Study of CTF for the OECD/NEA LWR Uncertainty Analysis in Modeling Benchmark

This work describes the results of a quantitative uncertainty analysis of the thermal-hydraulic subchannel code for nuclear engineering applications, Coolant Boiling in Rod Arrays-Three Field (COBRA-TF). CTF is used, which is a version of COBRA-TF developed in cooperation between the Consortium for Advanced Simulation of Light Water Reactors and North Carolina State University. Four steady-state cases from Phase II Exercise 3 of the Organisation for Economic Co-operation and Development/Nuclear Energy Agency Light Water Reactor Uncertainty Analysis in Modeling (UAM) Benchmark are analyzed using the statistical analysis tool, Design Analysis Kit for Optimization and Terascale Applications (Dakota). The input parameters include boundary condition, geometry, and modeling uncertainties, which are selected using a sensitivity study and then defined based on expert judgment. Here, a forward uncertainty quantification method with Latin hypercube sampling (LHS) is used, where the sample size is based on available computational resources.

42 ENGINEERING↗

Uncertainty Reduction using Bayesian Inference and Sensitivity Analysis: A Sequential Approach to the NASA Langley Uncertainty Quantification Challenge

This paper presents a computational framework for uncertainty characterization and propagation, and sensitivity analysis under the presence of aleatory and epistemic un- certainty, and develops a rigorous methodology for efficient refinement of epistemic un- certainty by identifying important epistemic variables that significantly affect the overall performance of an engineering system. The proposed methodology is illustrated using the NASA Langley Uncertainty Quantification Challenge (NASA-LUQC) problem that deals with uncertainty analysis of a generic transport model (GTM). First, Bayesian inference is used to infer subsystem-level epistemic quantities using the subsystem-level model and corresponding data. Second, tools of variance-based global sensitivity analysis are used to identify four important epistemic variables (this limitation specified in the NASA-LUQC is reflective of practical engineering situations where not all epistemic variables can be refined due to time/budget constraints) that significantly affect system-level performance. The most significant contribution of this paper is the development of the sequential refine- ment methodology, where epistemic variables for refinement are not identified all-at-once. Instead, only one variable is first identified, and then, Bayesian inference and global sensi- tivity calculations are repeated to identify the next important variable. This procedure is continued until all 4 variables are identified and the refinement in the system-level perfor- mance is computed. The advantages of the proposed sequential refinement methodology over the all-at-once uncertainty refinement approach are explained, and then applied to the NASA Langley Uncertainty Quantification Challenge problem.

Uncertainty↗