Random Field Modeling of CPT Data

TLDR

CPT data include cone tip resistance, side friction, and pore‑water pressure measurements. The study aims to generate a priori 1D stochastic soil models from extensive CPT soundings for application at similar sites. Using 143 CPT soundings from a homogeneous 18 km² area near Oslo, only cone tip resistance was analyzed as independent 1D realizations of a statistically homogeneous 3D random field to build the model. Analysis shows cone tip resistance follows a fractional Brownian motion fractal stochastic model, with parameters estimated by maximum likelihood.

Abstract

An extensive set of cone penetration tests (CPT) soundings are analyzed statistically to produce an \Ia priori\N 1D stochastic soil model for use at other similar sites. The data were collected by the Norwegian Geotechnical Institute at the site of a new airport just north of Oslo, Norway, and consists of 143 CPT soundings over an area of about 18 km² in a reasonably homogeneous soil mass. The CPT data consist of cone tip resistance, side friction, and pore-water pressure measurements. Only the cone tip resistance is considered in this study, with it being considered closest to a “point” property of the soil, and only the vertical variation is characterized. To perform the statistical analysis, the data sets are viewed as independent 1D realizations extracted from a statistically homogeneous 3D random field. Plots of various transformations of the data indicate that the cone tip resistance records are best represented using a fractal stochastic model corresponding to so-called fractional Brownian motion, and its parameters are estimated via maximum likelihood.