Estimation of neutronics parameter sensitivity to nuclear data in random sampling-based uncertainty quantification calculations

摘要

We propose a method to estimate sensitivity profiles of neutronics parameters with respect to nuclear data in random sampling-based uncertainty quantification calculations. The proposed method is tested to estimate sensitivity profiles of fast neutron systems criticalities. A high order effect in sensitivity profile estimation is found to be quite important, so a reverse sampling method is developed to mitigate the high order effect. With this reverse sampling method, detail energy group-wise sensitivity profiles can be estimated even though some fluctuations are observed in specific sensitivity profiles. Energy-integrated sensitivity profiles can be accurately calculated with the proposed method. With the estimated sensitivity profiles, partial uncertainties, that are neutronics parameters uncertainties induced by specific nuclear data uncertainties, are also calculated. Numerical tests reveal that the proposed method reproduces quite well the reference partial uncertainties. A simple and practical partial uncertainty estimation method, which only requires a covariance matrix between neutronics parameter and nuclear data, is also tested and assessed.