Physico-statistical approach to assess the core damage variability due to a total instantaneous blockage of SFR fuel sub-assembly

•Physico-statistical tool for SFR safety for Total Instantaneous Blockage accident.•0D/1D but realistic physical models to describe the phenomenological event tree.•Twenty-seven uncertain parameters identified to cover all realistic accidental transients.•Uncertainty propagation performed via a Monte-Carlo sampling.•Quantification of safety margins: 18.1% of cases above the safety criterion.

Within the framework of the generation IV Sodium Fast Reactors (SFR) R&D program of CEA (French commissariat à l’énergie atomique et aux énergies alternatives), the safety in case of accidents is assessed. These accidental scenarios involve very complex transient phenomena. To get round the difficulty of modelling them, only ‘Bounding’ (most damaging) accidental conditions have been up to now studied for the safety demonstration. These transients are simulated with very complex multi-physical codes (such as SIMMER) which nevertheless include some adjusted and not well known parameters and require a long CPU (process) time preventing their direct use for uncertainty propagation and sensitivity studies, especially in case of a high number of uncertain input parameters. To cope with these constraints, a new physico-statistical approach is followed in parallel by the CEA. This approach involves the fast-running description of extended accident sequences coupling analytical models for the main physical phenomena in combination with advanced statistical analysis techniques. The efficiency of the methodology for the reactor safety analysis is demonstrated here for one type of accident – the Total Instantaneous Blockage (TIB) – which involves an extended range of complex physical phenomena. From the establishment of the physical models describing the TIB phenomenology, 27 uncertain input parameters and their associated probability distributions are identified. A propagation of these input parameter uncertainties is performed via a Monte-Carlo sampling, providing probability distribution of TIB outputs. A quantification of safety margins is also deduced.