Bayesian-estimation-based method for generating fragility curves for high-fidelity seismic probability risk assessment

This paper presents a Bayesian-estimation-based seismic fragility assessment method that can significantly reduce the computational burden. The likelihood function and prior distribution used in the Bayesian estimation were enhanced. A probability density function of the seismic capacity was adopted as the likelihood function, although previous studies have generally used the fragility curve. Further, the prior distribution was assumed based on the seismic capacity obtained by incremental dynamic analysis. A sophisticated parametric study was additionally conducted, and we obtained information on the unknown parameters of the prior distribution to establish the initial values. Preliminary investigations show that these enhancements reduced the required number of seismic simulations to approximately 1/5 of that of the conventional Bayesian method. Then, the method was applied to a reinforced-concrete water intake system, which is a critical component of the core-cooling system in nuclear power plants (NPPs). The results revealed that the fragility curves determined using the proposed method with 12 repetitions of the simulations were comparable to those using a detailed method developed by the Atomic Energy Society of Japan with 500 repeated simulations. Therefore, we concluded that the proposed method could be applied to high-fidelity structures to substantially reduce the computational burden.