^{a}and Hyung-Jo Jung

^{b}

^{a}heekun.ju@kaist.ac.kr

^{b}hjung@kaist.edu

Fragility analysis of nuclear power plant structures (NPPs) and their components are typically performed under log-normal assumption of the curves for the simplicity. In this method, structural nonlinearity is considered using scale factor with its median and the logarithmic standard deviation. However, the scale factor method for the calculation of fragility curve cannot be always appropriate because the effect of nonlinearity can be not constant along different intensity of ground motion. In this research, fragility curves are calculated using sampling-based method to account for structural nonlinearity in more realistic method. To do this, probabilistic seismic analysis of a representative NPP structure is implemented using an advanced Latin Hypercube sampling-based method which can extend the sample size checking the convergence of results. A numerical model of the structure is modelled to represent concrete shear wall behavior which dominates the nonlinear behavior of NPP structures. Uncertainties in material properties of structure and input ground motion is included in the sampling process. Probabilistic distribution of structural response from the analysis is conventionally assumed to be log-normal, but it is represented both in parametric and non-parametric method for the comparison. The fragility curves calculated from the response densities are also represented in log-normal and non-parametric distribution. Both representation ways of fragility curves under the structural nonlinearity are compared and evaluated.