[2I01-09] PRA Method 2

10:30 AM - 10:45 AM JST(1:30 AM - 1:45 AM UTC)

[2I05] Development of AI surrogate models for seismic risk assessmentPart 2: Study of Training Data

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

Keywords:

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

The authors are developing AI surrogate models to perform seismic risk assessment with correlation in a fast and efficient manner. The improvement of AI surrogate model performance necessitates the preparation of appropriate training data. In this study, we focused on examining the training data to enhance the performance of AI surrogate models. We trained the AI surrogate model with completely uncorrelated data. Additionally, we prepared two types of data: uncorrelated data using the union and intersection formula from set theory, and partially correlated data using Monte Carlo simulations, and compared the analytical results of the AI surrogate model. This presentation reports on the evaluation of the generalization performance of the AI surrogate model based on these analyses.