The objective of this study was to develop an understanding of the seismic demand sensitivity of a reactor containment building (RCB) of a nuclear power plant with respect to uncertainty in the input parameters. For safety inspections and maintenance throughout the life of a RCB, risk, reliability, and fragility assessments are necessary. These factors are highly dependent on the seismic responses of the structure, which can vary due to the uncertainty of the input parameters. This demonstrates the need to perform a sensitivity analysis of the seismic demand of the RCB.

Peak floor displacement (PFD) and peak floor acceleration (PFA) were selected as engineering demand parameters in this study. The RCB’s seismic demand sensitivity was studied at five different locations. Considering the uncertainty of five input parameters, compressive strength of concrete , tensile strength of concrete, yield strength of steel, mass, and damping ratio , 200 samples of the RCB were generated and each sample was subjected to 10 ground motions (matched with the U.S. Nuclear Regulatory Commission 1.60 spectra). Hence, 2000 nonlinear time history analyses were performed.

Intermittent metamodels, Bayesian additive regression trees (BART), were constructed, and the performance and stability of the model were investigated using the data sets. Finally, Sobol’s sensitivity indices were estimated, employing the constructed BART model. It is reported that is the most sensitive parameter for PFA. Meanwhile, and are dominating parameters for PFD.