Improving the estimation of soil spatial variability by considering transformation uncertainty based on LDRFE analysis

Abstract

Transformation uncertainty is an imperative consideration in the process of transforming measurement data to desired design parameters in geotechnical engineering. In the reliability-based design framework, an improved understanding of transformation uncertainty is expected to produce more appropriate design of geotechnical structures. This paper quantifies the site-specific and aleatory transformation uncertainty and investigates how to consider it in estimation of random field parameters through numerically simulated cone penetration tests in spatially variable clay by means of large deformation random finite element analysis. The transformation uncertainty of a transformation model is modelled as a random field. Random field parameters of transformation uncertainty, including the standard deviation, scale of fluctuation and sample path smoothness, are estimated using maximum likelihood estimation along with Whittle-Matérn autocorrelation function model. Random field parameters of predicted shear strength with and without considering transformation uncertainty are thereafter compared with those random field parameters of actual shear strength. Results show that the consideration of transformation uncertainty in the transformation model can considerably improve the prediction accuracy of random field parameters. It is important to consider the spatial correlation structure of transformation uncertainty. This study offers a simple-to-use framework to estimate random field parameters considering the site-specific and aleatory transformation uncertainty.