Efficient global sensitivity analysis of structural vibration for a nuclear reactor system subject to nonstationary loading

•Non-stationary stochastic loads were applied to a reactor system finite element model.•Model parameters were varied based on variations amongst reactor designs, considering aging phenomena.•Reactor vessel head acceleration response was examined via global sensitivity analysis.•Kriging surrogates were trained by exercising the finite element model through several sampling strategies.•The resultant global sensitivity analysis and surrogate modeling errors were compared.•Efficient methodology enables parsimonious and credible design analysis techniques. The structures associated with the nuclear steam supply system (NSSS) of a pressurized water reactor (PWR) include significant epistemic and aleatory uncertainties in the physical parameters, while also being subject to various non-stationary stochastic loading conditions over the life of a nuclear power plant. To understand the influence of these uncertainties on nuclear reactor systems, sensitivity analysis must be performed. This work evaluates computational design of experiment strategies, which execute a nuclear reactor equipment system finite element model to train and verify Gaussian Process (GP) surrogate models. The surrogate models are then used to perform both global and local sensitivity analyses. The significance of the sensitivity analysis for efficient modeling and simulation of nuclear reactor stochastic dynamics is discussed.