Characterization of Nuclear Fuel using Multivariate Statistical Analysis

摘要

Various combinations of reactor type and fuel composition have been characterized using principle components analysis (PCA) of the concentrations of 9 U and Pu isotopes in the fuel as a function of burnup. The use of PCA allows the reduction of the 9-dimensional data (isotopic concentrations) into a 3-dimensional approximation, giving a visual representation of the changes in nuclear fuel composition with burnup. Real-world variation in the concentrations of 234U and 236U in the fresh (unirradiated) fuel was accounted for. The effects of reprocessing were also simulated. The results suggest that, even after reprocessing, Pu isotopes can be used to determine both the type of reactor and the initial fuel composition with good discrimination. Finally, partial least squares discriminant analysis (PSLDA) was investigated as a substitute for PCA. Our results suggest that PLSDA is a better tool for this application where separation between known classes is most important.