Optimized HSIC‐based tests for sensitivity analysis: Application to thermalhydraulic simulation of accidental scenario on nuclear reactor

Physical phenomena are commonly modeled by numerical simulators. Such codes can take as input a high number of uncertain parameters and it is important to identify their influences on the outputs via global sensitivity analysis (GSA). However, these be time consuming, which prevents GSA based classical Sobol' indices, requiring too many code simulations. To address this limitation, paper focuses Hilbert–Schmidt independence criterion (HSIC) proposes new goal-oriented algorithms optimize permuted HSIC-based tests for screening ranking purposes. Built upon sample inputs/outputs studied simulator, relies estimation p-value under hypothesis. These p-values estimated permutation method whatever size, but reliable large permutations prohibitive in practice. overcome this, several strategies proposed greedy estimate p-value, according final goal GSA. Three sequential thus proposed: oriented, ranking-screening oriented. tested compared analytical examples, before being applied thermalhydraulic use case simulating an accidental scenario nuclear pressurized water reactor. Their efficiency saving clearly demonstrated. Moreover, convergence study, made computationally tractable optimized algorithms, carried out assess robustness results case.