企画セッション: リスクと不確実性の定量化

[2A13-21-03] 3D spatial distribution estimation considering nonstationary in the depth direction and improvement of calculation efficiency by Kronecker product

*Yukihisa Tomizawa¹, Ikumasa Yoshida¹, Yu Otake² (1. Tokyo Citi University, 2. Tohoku University)

Keywords:Three-dimensional spatial distribution estimation, geotechnical property, Gaussian process regression, non-stationarity, Kronecker product

Three-dimensional spatial distribution estimation of the geotechnical property is performed on the measured data using Gaussian process regression, and a blind test is performed to examine the accuracy. When the non-stationarity of the standard deviation of the random field is considered in the depth direction, the improvement of estimation accuracy is confirmed than the stationary case. It is also shown that the calculation cost is greatly reduced by using Kronecker product when estimating the parameters of the random field by the maximum likelihood method.