Implementing a three-dimensional stochastic model of the subsurface in current geotechnical design methods

摘要

In practice nowadays most geotechnical designs are based on a load and resistance factor design (LRFD) approach. LRFD approaches are adapted in for example the Dutch NEN-standards and the new Eurocode standards. In current design procedures uncertainties in load-conditions and strength and deformation parameters can be accounted for by means of so called "partial factors" and characteristic values for geotechnical soil parameters. The choice of the partial factors depends on the type of geotechnical construction and load conditions, the type of soil parameter and the type of limit state which is investigated. The procedure for determining the characteristic value is based on the assumption that geotechnical soil parameters can be described by a normal distribution characterised by its mean, standard deviation and coefficient of variation. The standards in their current form do have some limitations. By using the variation coefficient as a measure of the spread of geotechnical parameters, the influence of the spatial variability and spatial distribution of the parameters on the geotechnical design is neglected. Uncertainties in the geometry of the subsurface model, i.e. the thickness and depth of the lithological layers, are not adequately addressed in the standards. Both spatial variability of soil parameters and geometry strongly depend on the geological conditions of the site. Both aspects can be incorporated in design calculations when 3D stochastic models of the subsurface are used as "input". In the present paper the differences between the simplified standard probabilistic approach and the advanced 3D-stochastic modelling approach are described and illustrated in a case study.