Stability improvement of noise analysis method in the case of random noise contamination for subcriticality measurements

Abstract

The performances of five noise analysis methods (Rossi-alpha, cross-correlation, Feynman-alpha, Feynman difference, and power spectral density) have been evaluated for subcriticality measurement of nuclear reactors, and two new approaches have been proposed to improve the stability of k eff estimations: (1) introduction of multiple detector signals and (2) new formulation of the Rossi-alpha method considering random noise contaminations. Modified thermal Godiva problem has been designed for the evaluations. The analysis was performed with fission count signals from the core of the Godiva model on three cases according to the level of random noise. The five noise analysis methods showed high accuracy of the estimated k eff when there was no contamination. In the case of low level contamination, they showed stable results with acceptable accuracy. However, in the case of high level contamination, although the Rossi-alpha results still showed high accuracy, the stability of the five noise analysis methods was significantly decreased. In order to reduce the instability for estimating the keff, two novel approaches were proposed: one is to design eight-detector model problem, and the other is to derive the advanced Rossi-alpha equation with consideration of random noise. In the eight-detector problem, one signal from the core is equally divided by three planes passing through the center of the model geometry and perpendicular to one another. It describes the situation using multiple detector signals to reduce the adverse impact of random noise contamination. On the other hand, the advanced Rossi-alpha formulation was used as the fitting curve equation for the evaluations. As a result, it was confirmed that the new approaches improved the stability of keff estimations.