企画セッション: リスクと不確実性の定量化

[2A13-21-04] Efficient Estimation of Limit State Probability by Adaptive Surrogate Model using Gaussian Process Regression

*Tomoka NAKAMURA¹, Ikumasa YOSHIDA¹, Yu OTAKE² (1. Tokyo City Univercity, 2. Tohoku University)

Keywords:Surrogate Model, Active Learning, Gaussian Process Regression, Learning Function, Limit State Probability, Importance Sampling

This study discusses the points for improving the AK-MCS proposed by Echard et al. (2011) which is an efficient method with active surrogate model of limit state function for calculating limit state probabilities. The proposed method introduces Importance Sampling without using design points, and is applied to an 8-dimensional consolidation settlement problem to calculate the limit state probability. It is confirmed that the probability is calculated with small number of function call.