Probabilistic characterization of two-dimensional soil profile by integrating cone penetration test (CPT) with multi-channel analysis of surface wave (MASW) data

In situ, laboratory, and geophysical tests are currently used in site characterization. These tests explore different parts of a site measuring different engineering properties at different resolutions or scales. The test results are then used to derive a design profile. In traditional approaches, the positions of boundaries between geological units are identified first, and the soil profile is divided into several layers. Constant engineering properties are assigned to each geological unit and the variabilities within each layer are ignored. To take the uncertainties into account, characteristic design values are assigned. There are no commonly accepted guidelines for choosing design values, however, which introduces additional subjective uncertainties. This paper proposes a probabilistic site characterization approach, based on Bayesian statistical methods, that allows a design profile involving uncertainty to be determined automatically. The derived soil profile is not modelled by uniform layers, but by random fields, which can be used directly in probabilistic analysis. The proposed approach is verified by a synthetic example, and further applied to a soft soil test site in Ballina, New South Wales, Australia, and compared with traditional approaches. The results show that by gradually incorporating more data into the Bayesian inversion, the uncertainty in the soil profile is greatly reduced.