Bayesian Inference for the Quantitative Analysis of Nonuniformly Distributed Uranium Radioactivity in a Waste Drum

Abstract

The characteristics of radioactive waste should be identified for free release or permanent disposal. Radioactive waste from nuclear power plants and fuel fabrication facilities is usually stored in a 200-L drum. A sampling procedure or nondestructive assay (NDA) is used to evaluate the properties of the waste. The distribution of the measured nuclides in the drum is difficult to know in advance, resulting in a radioactivity bias in NDA. This nonuniform radioactivity distribution is a critical point that makes it hard to evaluate waste characteristics. The most commonly used NDA method for drum scanning is a segment Gamma scanning (SGS) method or a tomographic Gamma scanning (TGS) method. The SGS method produces inaccurate results when radionuclides are nonuniformly distributed within a drum. The TGS method, on the other hand, can accurately analyze radioactivity in nonuniform situations. However, the complexity of the mechanical configuration and a time-consuming calibration procedure causes inconvenience in equipment operation. In this study, we developed a Bayesian inference model for quantitatively analyzing nonuniformly distributed uranium radioactivity in a waste drum from a single measurement in a simple mechanical setup. Measurements with three-dimensionally distributed uranium powder (UO2) in a waste drum show that the proposed method analyzes radioactivity on average about six times more accurately than the SGS method.