Criticality codes biases/uncertainties determination for fissile nuclear material transportation using different approaches

Criticality safety analysis regarding fissile nuclear material transportation or operationsrequires, among many aspects, the experimental validation of the criticality codes with the associatedcross-section libraries. The requirements for the criticality code experimental validation is todemonstrate the similarity between a selected set of critical experiments and the industrialconfiguration studied in order to determine biases and the associated uncertainties due tocalculations. The determination of biases and the associated uncertainties can be done by usingdifferent approaches.The approach currently used in France for addressing the bias and its associated uncertaintyis mainly based on the “expert judgement” and the knowledge of available experiments. This approachuses descriptive parameters (geometry, composition?) and some macroscopic calculated parametersto infer experiments potentially representative of a studied case and corresponding biases. Thenbiases of the reference experiments are transposed to the studied case depending of therepresentativity of the reference cases. This step can be eased doing a linear regression of keff versusthe parameter that best describes the configuration.An alternative way to study the similarity between selected experiments and the applicationcase is to use the statistical approach based on the Generalized Linear Least Square Method(GLLSM). This method allows propagating uncertainties in nuclear data and discrepancies betweencalculations and reference benchmarks for a selection of experiments to linearly adjust calculated keffvalues to reference values and therefore exhibit a bias due to nuclear data for a studied case.The aim of the paper is to compare these two methodologies on an Orano TN’s transport andstorage cask involving BWR fuel burnt at around 15 GWd/MTU. A first selection of experiments isdrawn using an expert judgement. Then this selection is restricted to experiments that are shown to bethe closest to the studied case regarding the “similarity” parameter calculated with the SCALE 6.2.1package. Both methodologies are then applied on the selected case and give comparable biases withregards to the uncertainties.