Year
2018
Abstract
Dry storage cask has become more of a semi-permanent solution rather than a temporary one for spent fuel, since the permanent storage solutions has proved elusive over recent decades. It was estimated that 4500 dry casks will be loaded until 2025. Due to the increased service life of casks and rapid growth of loaded casks, safety concerns about the integrity of internal structure are rising. No internal sensors were placed in the casks and these casks can’t be open for obvious reasons. Therefore, DOE has funded an integrated research project to solve the spent fuel cask internal structure monitoring problem with nondestructive evaluation methods.In this work, a spent fuel assembly structure fault diagnosis scheme based on logistic regression model is developed. Logistic regression is a technique borrowed by machine learning from statistics, it can analysis a dataset in which there are one or more independent variables that determine an outcome, and its outcome is measured with a dichotomous variable (in which there are only two possible outcomes), it was used to predict a binary outcome (True/False) given a set of independent variables. A Transnuclear 32 cask (TN-32) with 32 loaded spent fuel assemblies was reproduced with detailed into fuel rod level in MAVRIC (Monaco with Automated Variance Reduction using Importance Calculation) sequence of SCALE6.2.1. Various radiation signature (fuel gamma-ray, structure gamma-ray, neutron, neutron induced gamma-ray) on every cask surface were calculated with MAVRIC under normal and accident conditions (spent fuel assembly structure fail). These data were collected to train the logistic regression model. The trained models are subsequently applied to data calculated with various spent fuel assembly structure fault conditions for spent fuel integrity diagnosis. This proposed scheme is being evaluated feasibility to correct detect unusual shifts or damage to spent nuclear fuel assembly inside long-termstored dry storage casks.