Estimating spatial correlations under man-made structures on soft soils

The material point method is a finite element variant which allows the material, represented by a point-wise discretization, to move through the background mesh. This means that large deformations, such as those observed post slope failure, can be computed. By coupling this material level discretization to the spatial variability of the material generated by random fields and embedding this into a Monte Carlo framework, a new method called the Random Material Point Method (RMPM) has been developed. This method retains the advantages of the so-called random finite element method, that is, a risk based interpretation of the influence of spatial variability of the material properties, but additionally enables the effective modeling of large deformations to give a risk based interpretation of post-failure mechanisms. After a brief introduction to the RMPM methodology, the analysis of an idealized cohesion strain-softening clay slope is presented, which illustrates the influence of anisotropy of the material variability on the evolution of retrogressive slope failures.