Cost-effective design of long spatially variable soil slopes using conditional simulation

The three dimensional nature of soil spatial variability implies the need for 3D analysis of geotechnical structures. This paper presents the probabilistic analysis of long slopes such as levees and highway embankments, which are usually analysed unrealistically in plane strain, thereby ignoring the discrete failure mechanisms often encountered in practice. Conditional random fields of soil heterogeneity have been generated based on 3D Kriging, so that they match measurement data at borehole locations and honour the spatial correlations of the soil properties. A simple example involving the cost-effective design of an excavation in a 3D clay deposit has been investigated. It has been demonstrated that, by using conditional random fields within the random finite element method, more cost-effective geotechnical designs can be achieved while maintaining the same calculated reliability.