A New Approach for Nuclear Data Covariance and Sensitivity Generation

Abstract

Covariance data are required to correctly assess uncertainties in design parameters in nuclear applications. The error estimation of calculated quantities relies on the nuclear data uncertainty information available in the basic nuclear data libraries, such as the U.S. Evaluated Nuclear Data File, ENDF/B. The uncertainty files in the ENDF/B library are obtained from the analysis of experimental data and are stored as variance and covariance data. The computer code SAMMY is used in the analysis of the experimental data in the resolved and unresolved resonance energy regions. The data fitting of cross sections is based on generalized least‐squares formalism (Bayes’ theory) together with the resonance formalism described by R‐matrix theory. Two approaches are used in SAMMY for the generation of resonance‐parameter covariance data. In the evaluation process SAMMY generates a set of resonance parameters that fit the data, and, in addition, it also provides the resonance‐parameter covariances. For existing resonance‐parameter evaluations where no resonance‐parameter covariance data are available, the alternative is to use an approach called the “retroactive” resonance‐parameter covariance generation. In the high‐energy region the methodology for generating covariance data consists of least‐squares fitting and model parameter adjustment. The least‐squares fitting method calculates covariances directly from experimental data. The parameter adjustment method employs a nuclear model calculation such as the optical model and the Hauser‐Feshbach model, and estimates a covariance for the nuclear model parameters. In this paper we describe the application of the retroactive method and the parameter adjustment method to generate covariance data for the gadolinium isotopes.