Uncertainty analysis based on Bayesian inference for partial defect verification of PWR spent nuclear fuel

Abstract

Ensuring the integrity of spent nuclear fuel (SNF) is essential for nuclear non-proliferation efforts. While detecting gross defects is relatively straightforward, identifying partial defects remain challenging. This study proposes a Bayesian inference method implemented by our newly developed Yonsei Single-photon Emission Computed Tomography version 2 (YSECT.v.2) for verifying partial defects in SNF. Unlike traditional SNF defect detection algorithms that estimate specific values, the proposed method estimates distributions, thus providing belief in the estimates. Using the Monte Carlo (MC) method, we simulated partial defect scenarios and evaluated the proposed method's effectiveness against maximum-likelihood expectation-maximization (MLEM) across various defect patterns, ratios, and heterogeneous burnup conditions. The results indicate that the proposed technique reliably detects nuclear material diversion with high confidence.