NON-INTRUSIVE ALTERNATIVE TO GENERALIZED LINEAR LEAST-SQUARES METHODOLOGY FOR CRITICALITY SAFETY APPLICATIONS

The generalized linear least-squares (GLLS) methodology is a state-of-the-art approach used in the criticality safety community to determine the calculational bias for a given application based on the measured biases from a number of relevant criticality benchmark experiments. In its deterministic rendition, the analyst requires access to the sensitivity profiles of the calculated k_eff values for the experiments and the application with respect to all nuclear data, described by first-order derivatives of k_eff with respect to the multigroup nuclide-reaction-ratedependent cross sections. In the absence of an adjoint sensitivity capability, the calculation of these derivatives becomes computationally infeasible, thereby limiting the GLLS applicability to adjoint-enabled codes only. This manuscript proposes an alternative probabilistic approach which precludes the need for k_eff derivatives. The approach is based on the execution of the forward model with randomized cross-section perturbations. This manuscript introduces the theoretical basis for the proposed approach and compares its performance to a standard GLLS rendition under SCALE's TSURFER module.