Random field characterization of stress-normalised cone penetration testing parameters

Abstract

Random field modelling of soil variability allows significant statistical results to be inferred from field data; moreover, it provides a consistent framework for incorporating such variability in reliability-based design. Cone penetration testing (CPT) is increasingly appreciated because of its near-continuity and repeatability. Stress-normalised CPT parameters are included in widely used engineering procedures. Nonetheless, the results of variability analyses for these parameters are surprisingly limited. This paper attempts to characterise normalised cone tip resistance (qc1N) and friction ratio (FR) rigorously using a finite-scale weakly stationary random field model. It must be emphasised that inherent soil variability so determined strictly refers to the variability of the mechanical response of soils to cone penetration. The variability of soil response potentially depends on the failure mode (shear for sleeve friction or bearing for tip resistance) and most probably on the volume of soil influenced (averaging effect). To investigate spatial variability, 70 physically homogeneous CPT profiles were first identified from 304 soundings (subdivided into five regional sites) and subsequently assessed for weak stationarity using the modified Bartlett test. Only 40 qc1N profiles and 25 FR profiles were deemed sufficiently homogeneous from both physical and statistical considerations for the scales of fluctuations to be valid and for estimation of the coefficient of variation of inherent soil variability. The majority of the acceptable profiles were found in sandy soils. The remaining profiles are in fine-grained soils, with a few in intermediate soils. Trends in the estimated random field parameters indicate that qc1N is more strongly autocorrelated than FR, probably because qc1N is influenced by a larger volume of soil around the cone tip, and that the mechanical response of cohesionless soils to cone penetration is significantly more variable and erratic than that of cohesive soils. Comparison with literature data indicates that normalisation leads to a decrease in the scale of fluctuation for cone tip resistance and a reduction in the coefficient of variation. A tentative explanation is that normalisation tends to minimise systematic in situ effects that are explainable by physical causes.