Uncertainty estimation of minimum detectable activity through sparse spectrum using kernel-based Gaussian process regression

Abstract

To address the growing challenge of managing the radioactive waste generated during the decommissioning of nuclear power plants, one area under focus has been liquid radioactive waste management. This area concerns the importance of adhering to clearance levels and regulations, where Minimum Detectable Activity (MDA) plays a crucial role in determining whether liquid radioactive waste can be safely released into the environment. In this study, we obtain the sparse spectrum to calculate the uncertainty by means of kernel-based Gaussian Process Regression (GPR). After calculating the uncertainty, we applied count correction to the sparse spectrum to match the count with the long measurement spectrum. We then corrected the sparse spectrum to match that of the measurement spectrum collected over 3000 s for comparison. The results show that as the measurement time increases, the MDA converges to the true value. By applying kernel-based GPR and correcting the sparse spectrum, the uncertainty estimation of MDA achieves significantly improved accuracy, especially when the measurement time exceeds 60 s. With a measurement time of at least 60 s, a reliable MDA estimate can be obtained, allowing for more efficient and accurate MDA assessments in practical scenarios. The significance of this study lies in the fact that it offers a faster and more efficient method for estimating MDA, which can potentially enhance the management of radioactive waste and improve the monitoring of nuclear facilities.