Comparison of Sensitivity Analysis Techniques in Monte Carlo Codes for Multi-Region Criticality Calculations

摘要

Recently, sensitivity and uncertainty (S/U) techniques have been used todeterminethe area of applicability (AOA) of critical experiments used for code and data validation. These techniques require the computation of energy-dependent sensitivity coefficients for multiple reaction typesfor every nuclide in each system included in the validation. The sensitivity coefficients, as used for this application, predict the relative change in the system multiplication factor due to a relative change in a given cross-section data component or material number density. Thus, a sensitivity coefficient, S, for some macroscopic cross section, (Sigma), is expressed as S = (Sigma)/k (partial-derivative) k/(partial-derivative)(Sigma), where k is the effective neutron multiplication factor for the system. The sensitivity coefficient for the density of a material is equivalent to that of the total macroscopic cross section. Two distinct techniques have been employed in Monte Carlo radiation transport codes for the computation ofsensitivity coefficients. The first, and most commonly employed, is the differential sampling technique. The second is the adjoint-based perturbation theory approach. This paper briefly describes each technique and presents the results of a simple test case, pointing out discrepancies in the computed results and proposing a remedy to these discrepancies.