[1D13-17] External Event PRA 2

[1D16] Development of AI surrogate models for seismic risk assessment

(1) Proposal of AI surrogate model

*Ryoto Kobayashi¹, Yasuki Ohtori¹, Hitoshi muta¹ (1. TCU)

Keywords:Surrogate Model, Seismic Risk Assessment, Monte Carlo simulation, Damage correlation

In seismic PRA, the impact of damage correlations on the risk profile cannot be ignored, so an evaluation that takes damage correlations into account is indispensable. However, the sum-product formula, which assumes complete independence among random variables for system analysis, cannot be used in risk assessment considering damage correlations. Therefore, Monte Carlo simulations are indispensable for seismic risk assessment. However, conducting a large number of Monte Carlo simulations is time-consuming and labor-intensive, which has been a challenge in conducting such assessments. In this paper, we report the results of a study on the applicability and effectiveness of an AI-based surrogate model developed to reduce computational costs.