[2I01-09] PRA Method 2

10:15 AM - 10:30 AM JST(1:15 AM - 1:30 AM UTC)

[2I04] Efficient Creation of Fragility Curves Using AI

*Yu Takano¹, Auron Ryu Wiles¹, Ryoto Kobayashi¹, Yasuki Ohtori¹, Hitoshi Muta¹ (1. TCU)

Keywords:

Fragility Curve,Transfer Learning,Feedforward Neural Network,Monte Carlo Simulation,Nuclear power plant Risk Assessment

In earthquake-prone Japan, fragility curves showing the probability of structural damage from seismic activity are crucial for assessing nuclear power plant risks. However, nuclear power plants have many components. Creating fragility curves typically needs Monte Carlo simulations and is time-consuming. This study investigates simplifying the creation of fragility curves using transfer learning and feedforward neural networks (FNN). Analysis involves using 400 images overlaying seismic spectra on structural curves for transfer learning. The results obtained from transfer learning are used to train an FNN to estimate the values of Am and beta. These estimates created fragility curves, compared to those from Monte Carlo simulations. The AI-estimated fragility curves matched those from Monte Carlo simulations.