S.M. Woo, H. Kim, and S.S. Chirayath, “Influence of the spatial Pu variation for evaluating the Pu content in spent nuclear fuel using Support Vector Regression”, Annals of Nuclear Energy, 135 (2020).
The correlation of the amount of Pu produced in spent nuclear fuel to burnup (BU), cooling time (CT), initial U-235 enrichment (IE), and axial location (AX) is investigated by the Support Vector Regression (SVR) method. The AX parameter is a new one compared to the other studies reported in the literature. The regression coefficient (R2) and root mean square error (RMSE) values are used to determine the accuracies between the use of a four-parameter (BU, CT, IE, AX) and a three-parameter (BU, CT, IE) SVR model in the predicting the local Pu amount in the spent nuclear fuel. The R2 value for the four-parameter case (0.9996) is closer to unity (best case) than that for the three-parameter case (0.9776). The RMSE for the four-parameter case (0.0034) is less than of the three-parameter case (0.0243). The results of the SVR based machine learning analyses to predict the axial variation of Pu mass density in nuclear fuel show that accurate results are obtained from using the four-parameter case when compared to the original predictions using the SERPENT code. The correlation coefficients for BU, CT, IE, and AX for the Pu mass density variation are also evaluated. From the correlation analysis, it is observed that the most strongly correlated parameter with the Pu mass density is BU. The observations from this study show that the errors in Pu mass density axial variation predictions can be mitigated by considering the AX parameter along with the BU, CT and IE parameters.