ABSTRACT:

This paper presents some preliminary results from a comprehensive study on the application of the finite element method for reliability analysis in geotechnics. Some remarks on the stochastic nature of soil properties are given together with typical values for their stochastic modelling. The numerical procedure adopted (a point estimate method) is briefly described. A finite-element code is used to process the input variables and determine the input parameters for the performance function. The practicability of the suggested procedure is shown by solving an idealized tunnel excavation problem.

INTRODUCTION

In the past two decades the importance of reliability analyses and the stochastic nature of soil properties has been recognized in practical geotechnical engineering. Although the advantages of probabilistic tools are now generally accepted, there is still only a limited number of applications to practical problems documented in the literature. The reason being that closed form solutions are not available for most practical cases and numerical methods involve high computational efforts. However recent achievements in numerical analysis and the fast development of hardware have significantly reduced run times and pre- and post processing efforts so that a probabilistic approach on the based on deterministic finite element results seems to be a feasible way for the future and thus worthwhile to investigate.

STOCHASTIC NATURE OF SOILS

In general three major sources of uncertainties in soil profile modelling can be distinguished (Vanmarcke 1977). In this paper the natural heterogeneity or inherent uncertainty is dealt with only. In practice most geotechnical problems are still solved using a deterministic approach employing soil parameters which are Òon the safe sideÓ. In order to assess the behaviour of a geotechnical structure or the change of the deterministic factor of safety with respect to variable parameters such as e.g. shear strength sensitivity analyses are usually carried out.

This content is only available via PDF.
You can access this article if you purchase or spend a download.