PNNL

Abstract

Detecting impurity gases in spent nuclear fuel (SNF) canisters is a novel structural health monitoring (SHM) approach to an important problem. The SNF canisters are sealed containers that cannot be inspected visually. Acoustic sensing can be deployed by taking advantage of the gaps in the canister. While ultrasonic time-of-flight (TOF) measurement provides valuable information, it is limited to binary gases and is unable to quantify two impurity concentrations in a three-component gas mixture. In this case, deep learning algorithms, particularly Convolutional Neural Networks (CNNs), offer a promising solution for gas composition analysis. In this study, CNN-based deep learning models were implemented to detect and quantify impurity gases in helium. An experimental platform was established to simulate canister conditions, and ultrasonic test data were collected. The data set considered the presence of argon and air in helium at concentrations ranging from 0% to 1.2% with at increments of 0.05%. Results showed that the multi-layer perceptron (MLP), decision tree (DT), and logistic regression (LR) classifiers achieved high accuracies in differentiating the acoustic responses travelling through pure helium versus contaminated helium. Furthermore, the ensemble CNN model exhibited improved concentration predictions and the ability to balance individual gas concentration by integrating 1-D and 2-D CNN models. These findings contribute innovative solutions for impurity gas detection and analysis within SNF canisters, thus ensuring safe storage and management of SNFs.