Engineering Papers⌕ Search

Engineering topics

Palmer, Todd S.

Publications and source records attributed to Palmer, Todd S..

At least 19 records

A variance deconvolution estimator for efficient uncertainty quantification in Monte Carlo radiation transport applications

Monte Carlo simulations are at the heart of many high-fidelity simulations and analyses for radiation transport systems. As is the case with any complex computational model, it is important to propagate sources of input uncertainty and characterize how they affect model output. Unfortunately, uncertainty quantification (UQ) is made difficult by the stochastic variability that Monte Carlo transport solvers introduce. The standard method to avoid corrupting the UQ statistics with the transport solver noise is to increase the number of particle histories, resulting in very high computational costs. In this contribution, we propose and analyze a sampling estimator based on the law of total variance to compute UQ variance even in the presence of residual noise from Monte Carlo transport calculations. Here, we rigorously derive the statistical properties of the new variance estimator, compare its performance to that of the standard method, and demonstrate its use on neutral particle transport model problems involving both attenuation and scattering physics. We illustrate, both analytically and numerically, the estimator’s statistical performance as a function of available computational budget and the distribution of that budget between UQ samples and particle histories. We show analytically and corroborate numerically that the new estimator is unbiased, unlike the standard approach, and is more accurate and precise than the standard estimator for the same computational budget.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Validation of time-dependent shift using the pulsed sphere benchmarks

The detailed behavior of neutrons in a rapidly changing time-dependent physical system is a challenging computational physics problem, particularly when using Monte Carlo methods on heterogeneous high-performance computing architectures. A small number of algorithms and code implementations have been shown to be performant for time-independent (fixed source and k-eigenvalue) Monte Carlo, and there are existing simulation tools that successfully solve the time-dependent Monte Carlo problem on smaller computing platforms. To bridge this gap, a time-dependent version of ORNL’s Shift code has been recently developed. Shift’s history-based algorithm on CPUs, and its event-based algorithm on GPUs, have both been observed to scale well to very large numbers of processors, which motivated the extension of this code to solve time-dependent problems. The validation of this new capability requires a comparison with time-dependent neutron experiments. Lawrence Livermore National Laboratory’s (LLNL) pulsed sphere benchmark experiments were simulated in Shift to validate both the time-independent as well as new time-dependent features recently incorporated into Shift. A suite of pulsed-sphere models was simulated using Shift and compared to the available experimental data and simulations with MCNP. Overall results indicate that Shift accurately simulates the pulsed sphere benchmarks, and that the new time-dependent modifications of Shift are working as intended. Validated exascale neutron transport codes are essential for a wide variety of future multiphysics applications.

Palmer, Camille J.↗

A Monte Carlo Thermal Radiative Transfer Solver with Nonlinear Elimination

Here in this paper, we present a new Monte Carlo method for solving the thermal radiative transfer (TRT) equations via the method of nonlinear elimination (NLEM). This method is inspired by the previous application of NLEM to thermal radiation diffusion. Our approach, called diffusion accelerated Implicit Monte Carlo (DAIMC), is a hybrid technique which combines a Monte Carlo method for solving a purely-absorbing transport equation and a diffusion solution that accounts for effective scattering, or absorption–reemission. The method aims to improve the implicitness of the traditional implicit Monte Carlo (IMC) method. We derive DAIMC generally for 3D Cartesian geometries, but in this paper, we present results and analysis in 1D slab geometry. These preliminary results indicate that DAIMC implementations may provide more accurate and robust TRT solutions than IMC in certain test problems.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Accelerated Deterministic Phonon Transport With Consistent Material Temperature and Intensities

Abstract We present a method for deterministically solving the frequency and temperature dependent phonon radiative transport (PRT) equation in the single-mode relaxation time (SMRT) approximation in the self-adjoint angular flux (SAAF) form. To handle the nonlinear coupling between the phonon intensities and the material temperature, we apply a linearization approach that is similar to one in thermal radiative transport. This procedure leads to the PRT equation with pseudo-scattering. The method presented includes acceleration of both the inner pseudo-scattering source iterations and outer temperature iteration with a gray diffusion synthetic acceleration (DSA) and Anderson acceleration, respectively. We use the finite-element method to discretize the PRT equation in space and the method of discrete ordinates (SN) for angular discretization. The proposed method is verified by a gray method of manufactured solutions problem and demonstrated on a problem using temperature and direction dependent multigroup data from lithium aluminate (LiAlO2). The iterative performance of the acceleration method in each test is then compared to the unaccelerated method.

Engineering↗

Proper orthogonal decomposition based reduced-order modeling of flux-Limited gray thermal radiation

Here, in this work, a proper orthogonal decomposition (POD) based reduced-order model (ROM) is developed to solve gray, flux-limited thermal radiation diffusion. We focus on the variable opacity radiation penetration benchmark posed by Olson, Auer, and Hall. The T -3 relationship for opacity in conjunction with high-temperature radiation penetrating an initially cold material produces a strong thermal radiation shock. This class of problems is particularly challenging for standard POD-based reduced-order modeling due to the nonlinearities presented by 1) the T 4 source term, and 2) flux-limited diffusion operator. To address these challenges and develop a cost competitive ROM, we employ a “hyper-reduction” technique through discrete empirical interpolation (DEIM) and allow for adaptive reduced-order projections through principal interval decomposition (PID). Performance of the proposed methodology is quantified by comparing the cost savings and accuracy relative to a full-order computation. Reference solutions and snapshot data are obtained through a full-order calculation performed by the University of Chicago maintained astrophysics code, FLASH. For consistency and potential extensibility, the developed ROM is also implemented in FLASH. We find that in the initialization regime, where the thermal radiation wave is initially created by the warming of the material, this class of problems is highly reducible and suitable for POD-based ROMs. However, the strong convective nature of the wave propagation regime is less reducible and more challenging to create an efficient ROM.

42 ENGINEERING↗

Correlation Calculations for the Russian Pu Metal Fast Experiments

Nuclear criticality experiments are often conducted in campaigns with multiple variations. These experiments reuse the same basic components, like the fuel, moderator, or positioning machines. The components have uncertainties in their geometry and composition that propagate to models of the experiments. Shared components create shared uncertainty between the $k_{eff}$ of benchmarks. The shared uncertainty is commonly quantified with a covariance, or correlation coefficient. These covariances can impact criticality safety and nuclear data validation applications. While benchmark evaluations tabulate an experiment’s uncertainty, they often lack a detailed calculation of correlations between experiments. Even some very commonly used benchmarks, like the Russian Pu Metal Fast (PMF) experiments, have missing correlations. This paper presents our approach to calculate the correlations for five of the Russian PMF experiments. The experiments share hemispherical Pu shells that induce a correlation between modeled $k_{eff}$ values. We estimated the correlations with simplified and detailed models of the experiments through linear-perturbation theory. The correlations between the experiments vary significantly between the detailed vs. simplified models. We also investigate how the correlations affect validation metrics of the experiments.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

An agent-based blackboard system for multi-objective optimization

In the field of multi-objective optimization, there are a multitude of algorithms from which to choose. Each algorithm has strengths and weaknesses associated with the mechanics for finding the Pareto front. Recently, researchers have begun to examine how multi-agent environments can be used to help solve multi-objective optimization problems. In this work, we propose a multi-objective optimization algorithm based on a multi-agent blackboard system (MABS). The MABS framework allows for multiple agents to read and write pertinent optimization problem data to a central blackboard agent. Agents can stochastically search the design space, use previously discovered solutions to explore local optima, or update and prune the Pareto front. A centralized blackboard framework allows the optimization problem to be solved in a cohesive manner and permits stopping, restarting, or updating the optimization problem. The MABS framework is tested against three alternative optimization algorithms across a suite of engineering design problems and typically outperforms the other algorithms in discovering the Pareto front. A parallelizability study is performed where we find that the MABS is able to evaluate a set number of designs, which require an evaluation time ranging from 0 to 300 seconds, quicker than a traditional optimization algorithm: this fact becomes more apparent the longer it takes to evaluate a design. To provide context for the benefits provided by MABS, a real-world nuclear engineering design problem is examined. MABS is used to examine the placement of experiments in a nuclear reactor, where we are able to evaluate hundreds of configurations for experimental placement while maintaining a strict set of safety constraints.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Introduction to an MCGIDI Mini-App and Performance Comparisons with XSBench

In high-performance Monte Carlo (MC) radiation transport codes, cross section lookups often account for the majority of computational expens. In response to this observation, a number of nuclear data mini-apps have been developed to profile and optimize the lookup process. A review of these mini-apps can be found in Ref. In contrast to a fully fledged MC code, nuclear data mini-apps do not feature particle tracking; instead, the cross section lookup process is considered in isolation, a simplification that allows for easier implementation and evaluation of various lookup schemes and techniques. GIDI+, as described on its public repository1, is ...a collection of C++ libraries for accessing evaluated and processed nuclear data stored in the Generalized Nuclear Database Structure (GNDS). In addition to reading GNDS files, GIDI+ has functions to sum and collapse multi-group data as needed by deterministic transport codes, and to sample GNDS data as needed by Monte Carlo transport codes. Certain modes of Mercury, a LLNL-developed MC radiation transport code, utilize GIDI+ for cross section lookups. Specifically, Mercury uses the MCGIDI library contained within GIDI+ to perform lookups on GPUs. A mini-app exercising MCGIDI’s cross section lookup capabilities would prove useful for optimization, which in turn could improve the performance of client codes such as Mercury. Such a mini-app could also be used to assess the performance of MCGIDI compared to other nuclear data mini-apps. In this document, we introduce a nuclear data mini-app built with the MCGIDI library. The MCGIDI mini-app (MCGIDI-MA) was developed to closely resemble the methodology present in XSBench, a nuclear data mini-app developed at Argonne National Laboratory. The remainder of this section provides an overview on XSBench. Section 2 describes MCGIDI-MA and its usage. Section 3 contains performance results of MCGIDI-MA on LLNL’s Lassen and Quartz compute platforms, as well as comparisons to XSBench performance where appropriate. In Section 4, we present a summary of our findings.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Fully Implicit Monte Carlo (FIMC) (September 2019-September 2020 Progress Report)

This research is focused on applying the NLem method to the full thermal radiative transfer (TRT) equations in the context of a Monte Carlo solution strategy. A framework has been developed for obtaining a solution to the TRT equations using the nonlinear elimination method (NLem) method. Current efforts are focused of the development and verification of the required Monte Carlo algorithm, diffusion algorithm, and material energy solver. Once all three pieces have been verified, the algorithms will then be coupled into a fully functioning method for applying NLem to the TRT equations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Application of Machine Learning Algorithms to Identify Problematic Nuclear Data

In this work we aim to show that Machine learning algorithms are promising tools for the identification of nuclear data that contribute to increased errors in transport simulations. We demonstrate this through an application of a machine learning algorithm (Random Forest) to the Whisper/MCNP6 criticality validation library to identify nuclear data that are associated with an increase of the bias (simulated - experimental $k_{eff}$) in the calculations. Specifically, the $k_{eff}$ sensitivity profiles (w.r.t. nuclear data) of 233 U solution benchmarks are used to predict the bias and Shapley Additive Explanations (SHAP) are used to explain how the sensitivities are related to the predicted bias. The SHAP values can be interpreted as sensitivity coefficients of the machine learning model to the $k_{eff}$ sensitivities which are used to make predictions of bias. Using the SHAP values we can identify specific subsets of nuclear data which have the highest probability of influencing bias. We demonstrate the utility of this method by showing how SHAP values were used to identify an inconsistency in the 19 F inelastic scattering nuclear data. The methodology presented here is not limited to transport problems and can be applied to other simulations if there are experimental measurements to compare against, simulations of those experimental measurements, and the ability to calculate sensitivities of the model output with respect to the data inputs.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Generation of Multigroup Multiband Nuclear Data

Historically, deterministic pre-processing methods have been used for generating multigroup (MG) and multigroup multiband (MGMB) cross sections. These deterministic methods require an assumption of the shape for the scalar ux. There has been increasing interest in utilizing Monte Carlo transport as a method to generate these cross sections since the energy self-shielding and resonance overlap are fully accounted for. The purpose of this project is to implement and test the capability of generating these cross sections in Mercury.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Prediction of thermal conductivity in dielectrics using fast, spectrally-resolved phonon transport simulations

We present a new method for predicting effective thermal conductivity in materials, informed by ab initio material property simulations. Using the Boltzmann transport equation in a self-adjoint angular flux formulation, we performed simulations in silicon at room temperatures over length scales varying from 10 nm to 10 micron and report temperature distributions, spectral heat flux and thermal conductivity. Our implementation utilizes a Richardson iteration on a modified version of the phonon scattering source. In this method, a closure term is introduced to the transport equation which acts as a redistribution kernel for the total energy bath of the system. This term is an effective indicator of the degree of disorder between the spectral phonon radiance and the angular phonon intensity of the transport system. In conclusion, we employ polarization, density of states and full dispersion spectra to resolve thermal conductivity with numerous angular and spatial discretizations.

36 MATERIALS SCIENCE↗

Diffusion Synthetic Acceleration for High Order S{sub N} Transport on Meshes with Curved Surfaces

Thermal radiation transport is integral to multiphysics applications such as inertial confinement fusion and astrophysics. Materials in these problems can be exceptionally optically thick and diffusive. A transport solver must be capable of accurate and efficient solutions to such problems. Previously, Woods et al. established a proof-of-concept for high order (HO) discontinuous Galerkin finite element (DGFEM) transport on meshes with curved surfaces. They numerically demonstrated the spatial convergence is order p + 1. Remarkably, these convergence rates are relatively invariant to the amount of curvature of the spatial mesh. We present here an extension of this work that demonstrates a rapidly-convergent iterative solver for the FEM transport equations on meshes with curved sides. We use the modified interior penalty (MIP) diffusion synthetic acceleration (DSA) established by Wang and Ragusa. The partially consistent MIP equations stem from the symmetric interior penalty (IP) derivation of the diffusion equation. The IP method is not stable for optically thick cells so it was combined with the diffusion conforming form (DCF), a discretized diffusion equation derived from the discretized S N equations that is not stable for intermediate optical thicknesses. The MIP equations adapt a 'switch' to the IP method in optically thinner regions and the DCF method in optically thicker regions. The literature does not reveal any Fourier analyses performed for iterative transport on meshes with curved surfaces. Wang and Ragusa performed a Fourier analysis for orthogonal meshes and our results appear to be as expected. Previous literature has reported on some of the sensitivities that affect the spectral radius of DSA schemes: polynomial order, and scattering ratios. Presently, we focus on the sensitivity of the spectral radius to polynomial order, scattering ratio, the constant C, and mesh curvature for varying cell thicknesses. We also numerically study the behavior of the spectral radius to verify our DSA implementation. A sequence of problems is designed to exercise the acceleration. As the problems become more optically thick and diffusive, the accelerated solution continues to converge toward the analytic solution. We employ the Modular Finite Element Methods (MFEM) package, developed at Lawrence Livermore National Laboratory, to generate the linear algebraic system of equations for each angle in the discrete ordinates quadrature set, and use UMFPack to solve this system of equations.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Semi-Analytical Benchmarks for MCNP6

Code verification is an extremely important process that involves proving or disproving the validity of code algorithms by comparing them against analytical results of the underlying physics or mathematical theory on which the code is based upon. Monte Carlo transport codes such as MCNP6 undergo verification and testing upon every release to ensure that the codes are properly simulating nature. Specifically, MCNP6 has multiple sets of problems with known analytic solutions that are used for code verification. Most of the verification problems are fairly simple due to the fact that they are usually composed of few materials, with simplified cross-sections, or physics approximations, and simple geometry. These approximations make the transport equation possible to solve analytically or semi-analytically, and thus they are quite valuable in verifying that the code algorithms and methods (Monte Carlo, discrete ordinates, etc.) are operating as intended. The remainder of this paper is organized as follows: first we briefly discuss the theory behind the benchmarks problems, then the numerical results are presented, and finally we conclude with general remarks on this and possible future work.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Semi-Analytical Benchmarks for MCNP6 [Slides]

This presentation begins with an Introduction and some background on semi-analytical benchmarks for MCNP6. Discussion topics include: verification & validation, MCNP history of V&V, semi-analytic benchmarks, numerical Results, comparison of semi-analytic to MCNP, and tips for proper comparison. This presentation ends with conclusions & future Work.

97 MATHEMATICS AND COMPUTING↗

Quantifying the Uncertainty in Deterministic Phonon Transport Calculations of Thermal Conductivity using Polynomial Chaos Expansions

The nature of computational simulations requires the inclusion of an uncertainty analysis, as we have limited knowledge of all physically determined input parameters for a computationally simulated problem. We rely on uncertainty quantification (UQ) to characterize our confidence in the outcomes. Quantifying uncertainty can provide a basis for certifications in high-consequence decisions, such as nuclear reactor design, and is a fundamental component of model validation. We employ a previously developed method of uncertainty quantification, polynomial chaos expansion with stochastic collocation (PCE-SC), applied to deterministic phonon transport simulations. In these simulations, we use the neutron transport code Rattlesnake, which solves the Self-Adjoint Angular Flux (SAAF) formulation of the transport equation with a continuous finite element (CFEM) spatial discretization and discrete ordinates, spherical harmonics angular discretizations. Rattlesnake was developed in the multi-physics object oriented simulation environment (MOOSE) framework. We have previously shown Rattlesnake to be effective in simulating phonon transport. We aim to provide a deterministic phonon transport framework for heterogeneous nuclear fuel with fission product defects to predict thermal conductivity (?). A first principles, physics-based calculation of thermal conductivity must involve factors such as the microstructure of nuclear fuel, which constantly changes during the fission process through the formation of isotopic decay products. Heat transport in oxide nuclear fuels is dominated by phonon transport. Impurities in the bulk material influence the transport of energy at the fundamental level, altering the scattering behavior of phonons and electrons. Conventionally, heat transport follows classical physics based on the heat equation derived from Fourier's law q = -k∇T, where q is heat flux, k is thermal conductivity and ∇T is a temperature gradient. However, Fourier's law is a macroscopic empirical law in which the thermal conductivity? does not have a mechanistic connection to the underpinning heat transport processes. Thermal conductivity of a material depends both on the material's intrinsic ability to transport heat and a variety of resistive effects caused by defects in the material. Thus ab initio prediction of the macroscopic conductivity - the property of interest for safe reactor operation - requires simulating detailed processes of heat transport, and then determining the effective thermal conductivity of the simulated material from the resulting heat flux under the imposed temperature difference. The thermal conductivity of a bulk, homogeneous, dielectric material can be well estimated by the mechanistically derived expression: k{sub bulk} = 1/3 C{sub ν}νgλ (2), where C{sub ν}, νg, and λ are the volumetric specific heat, phonon speed, and phonon mean free path, respectively. In these problems the relevant parameter is the material's acoustic thickness, its characteristic distance L relative to the phonon mean free path λ. In Eq. (2) the largest source of uncertainty is λ. While propagating uncertainty in λ to uncertainty in k through Eq. (2) is trivial, systems with extrinsic scattering require more sophisticated approaches. At the microscopic level, uncertainty in λ changes both a material's intrinsic thermal conductivity. Our goal is to propagate the uncertainty in λ through the deterministic transport computation of k. Because there are a small number of uncertain input quantities, we use the method of polynomial chaos expansion (PCE), which expresses solutions in the form of spectral expansions of an uncertain variable. This approach combines both intrusive and nonintrusive methods of uncertainty propagation techniques and results in a unique formulation which is very effective and efficient for problems with few uncertain parameters. Uncertainty Quantification and Sensitivity Analysis We use PCE-SC to measure propagation of uncertainty in a 3-D phonon transport problem of homogeneous silicon, to compute the mean and variance in temperature, heat flux and thermal conductivity. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

High Order Finite Elements S{sub N} Transport in X-Y Geometry on Meshes with Curved Surfaces

Radiation hydrodynamics is useful to model astrophysics, ICF, and glass manufacturing. A large amount of research has been conducted to understand how to solve hydrodynamics equations using high-order finite elements. Researchers at LLNL have developed a general high-order FEM library, Modular Finite Element Methods (MFEM), to enable this research. The BLAST code uses MFEM to evolve the Lagrangian hydrodynamics equations, resulting in meshes with higher order curved surfaces. Our research focuses on the requirements for high order finite element solutions of the radiation transport equation with the eventual goal of studying the complete high-order radiation hydrodynamics solution methodology. For this research we used MFEM to develop a proof-of-concept code that solves the high order finite element transport equations in Cartesian coordinates on meshes with curved surfaces. We have solved several test problems to assess the behavior of our method. (authors)

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

STEADY STATE MODELING OF THE MINIMUM CRITICAL CORE OF THE TRANSIENT REACTOR TEST FACILITY

With the advent of next generation reactor systems and new fuel designs, the U.S. Department of Energy (DOE) has identified the need for the resumption of transient testing of nuclear fuels. The DOE has decided that the Transient Reactor Test Facility (TREAT) at Idaho National Laboratory (INL) is best suited for future testing. TREAT is a thermal neutron spectrum, air-cooled, nuclear test facility that is designed to test nuclear fuels in transient scenarios. These specific scenarios range from simple temperature transients to full fuel melt accidents. DOE has expressed a desire to develop a simulation capability that will accurately model the experiments before they are irradiated at the facility. It is the aim for this capability to have an emphasis on effective and safe operation while minimizing experimental time and cost. The multi physics platform MOOSE has been selected as the framework for this project. The goals for this work are to identify the fundamental neutronics properties of TREAT and to develop an accurate steady state model for future multiphysics transient simulations. In order to minimize computational cost, the effect of spatial homogenization and angular discretization are investigated. It was found that significant anisotropy is present in TREAT assemblies and to capture this effect, explicit modeling of cooling channels and inter-element gaps is necessary. For this modeling scheme, single element calculations at 293 K gave power distributions with a root mean square difference of 0.076% from those of reference SERPENT calculations. The minimum critical core configuration with identical gap and channel treatment at 293 K resulted in a root mean square, total core, radial power distribution 2.423% different than those of reference SERPENT solutions.

Critical↗