Efficient estimation of structural response distribution for nonlinearly behaved nuclear power plant structures

Rigorous estimation of structural response distribution is essential for effective risk assessment of nuclear power plant (NPP) structures. The conventional approach for response estimation through probabilistic analysis involves conducting more than 30 analyses based on Latin Hypercube sampling (LHS) method to incorporate uncertainties, assuming the response follows a lognormal distribution. However, this approach can encounter limitations, such as non-convergence due to insufficient sample sizes or incompatibility with the lognormal assumption, especially in structures exhibiting nonlinear behavior—issues that can challenge less-experienced researchers. This paper proposes an efficient and stable method for estimating structural response distributions, overcoming these difficulties by adopting an advanced LHS method and a flexible distribution fitting model. The proposed method was applied to obtain estimates of the in-structure response spectrum (ISRS) for an NPP structure with hysteretic nonlinear behavior, demonstrating improved efficiency and a more precise statistical representation of the ISRS compared to conventional methods.