Masters Theses
Date of Award
5-2024
Degree Type
Thesis
Degree Name
Master of Science
Major
Nuclear Engineering
Major Professor
Vladimir Sobes
Committee Members
Sandra Bogetic, J. Wesley Hines, Jason Hayward
Abstract
Nuclear cross sections are a set of parameters that capture probability information about various nuclear reactions. Nuclear cross section data must be experimentally measured, and this results in simulations with nuclear data-induced uncertainties on simulation outputs. This nuclear data-induced uncertainty on most parameters of interest can be reduced by adjusting the nuclear data based on the results from an experiment. Integral nuclear experiments are experiments where the results are related to many different cross sections. Nuclear data may be adjusted to have less uncertainty by adjusting them to match the results obtained from integral experiments. Different integral experiments will adjust cross section data differently and reduce their associated uncertainties differently.
Various nuclear experts use simulation systems to calculate parameters of interest for various nuclear systems. These simulated application systems can make use of adjusted nuclear to reduce the uncertainties on their parameters of interest. Thus, different integral experiments have different impacts on reducing the uncertainties on the parameters of interest for different application systems.
This thesis verifies a method for predicting the uncertainty reduction on a parameter from a given integral experiment. This will allow the designers of integral experiments to predict how valuable their experiments will be for various applications and parameters. The verification is performed by first simulating an experiment system, an application system, the different possible cross section values that really exist, an imperfect measurement of the experiment system, the process of adjusting the cross section values, and the application of the adjusted values to the simulation of the application system. The results of that procedure are then compared to the uncertainty reduction predicted by a recently developed formula called the Uncertainty Reduction Ratio (URR) formula. The advantage of the URR formula is that it can be applied to a wider range of systems and parameters than those that had to be assumed for the verification procedure.
After verification of the URR formula, this thesis goes on to use the formula to demonstrate the optimization of the design of an integral experiment system for the reduction of uncertainty on parameters of interest for application systems. The experimental system whose optimization is being demonstrated is the Flexible Neutron Source (FNS) at the University of Tennessee Knoxville. The nature of the FNS optimization meant that many existing optimization algorithms for integral experiment design would be either inappropriate or inefficient.
The FNS a multi-purpose sub-critical experiment system planned to be constructed in the basement of the newly constructed Zeanah Engineering Complex. The FNS is a highly modular system that can be configured in an extremely large number of ways by rearranging a number of plates of various materials. The FNS is to be composed of 74 aluminum drawers (cassettes) arranged into three 5 by 5 layers (also referred to as zones), each of which is considered a distinct zone. Each cassette has room for twenty 0.5-inch material plates. These cassettes can be filled with plates made of a variety of materials depending on the application the FNS is being leveraged for and can be rearranged to perform different experiments. The number of combinations of plates in the FNS is large (~4470). A search algorithm needs to be used to determine a configuration that will be near optimal for adjusting cross section data for any given application.
A new edition of a general-purpose global optimization software package, Gnowee, is developed called Gnowee_multi. The new edition is improved in many ways including but not limited to: radical speed increases, parallelization, multi-objective capability, crash avoidance, crash recovery, and conversion to Python 3. The new edition is applied to the optimization of the FNS and sufficiently demonstrated its utility.
A demonstration optimization of the FNS was performed for an application reducing the uncertainty on the k-eigenvalue of a Sodium Fast Reactor (SFR). The optimization considered a limited design space, a limited set of unadjusted ENDF/B-VII.1 nuclear data, a very limited model of the measurement of the FNS, and calculated the sensitivities of the FNS (required for the URR formula) using the now out-of-date PERT MCNP 6.3 card (with associated limitations). Under those conditions, the final optimized designs suggested that the FNS could be used to reduce the uncertainty on the k-eigenvalue (A.K.A “the level of criticality”) of a generic SFR by between 15% and 20%.
The future work section describes extensions and variations of the URR formula as well as the replacement of the PERT card.
Recommended Citation
DePillis, Alexander Amedeo, "Genetic Algorithm Optimization of Experiment Design for Targeted Uncertainty Reduction. " Master's Thesis, University of Tennessee, 2024.
https://trace.tennessee.edu/utk_gradthes/11376
Included in
Artificial Intelligence and Robotics Commons, Nuclear Engineering Commons, Numerical Analysis and Scientific Computing Commons, Theory and Algorithms Commons