Generalized sensitivity analysis of responses to nuclear data in 2-D/1-D transport code KYADJ

Abstract

Nuclear data are important input parameters in reactor physics design. The uncertainty of the reactor calculation mainly originates from nuclear data variance. The approaches to evaluate the uncertainty are divided into the statistics sampling method and the “Sandwich” rule. The “Sandwich” rule is suitable for problems with plenty of input parameters and the key step of this method is sensitivity analysis. Sensitivity analysis includes the Forward Sensitivity Analysis Procedure (FSAP) and the Adjoint Sensitivity Analysis Procedure (ASAP). FSAP is simple to apply, but the increase of input parameters will cause a large number of repeated calculations. At present, ASAP based on the Classical Perturbation Theory (CPT) and the Generalized Perturbation Theory (GPT) is applied to various neutronics codes. GPT requires the solution of the generalized adjoint equations for responses. It is difficult for many neutronics codes to solve the generalized adjoint equation. In this paper, sensitivity analysis is developed in the two dimensions (2-D) and one dimension (1-D) coupled transport code KYADJ. KYADJ adopts the Forward Sensitivity Analysis Procedure based on the Reduced-Order Modeling (FSAP-ROM) to carry out the generalized sensitivity analysis. FSAP-ROM is tested by the PB-2 BWR pin cell benchmark and the SF96 lattice model. The error is verified by results from the Reactor Monte-Carlo (RMC) code and the direct forward perturbation calculations.