Multiscale stochastic finite element method on random field modeling of geotechnical problems — a fast computing procedure

Abstract

The Green-function-based multiscale stochastic finite element method (MSFEM) has been formulated based on the stochastic variational principle. In this study a fast computing procedure based on the MSFEM is developed to solve random field geotechnical problems with a typical coefficient of variance less than 1. A unique fast computing advantage of the procedure enables computation performed only on those locations of interest, therefore saving a lot of computation. The numerical example on soil settlement shows that the procedure achieves significant computing efficiency compared with Monte Carlo method.