Marine soil behaviour classification using piezocone penetration test (CPTu) and borehole records

Abstract

Several piezocone penetration test (CPTu)-based soil behaviour classification systems (SBCs) have been developed for standard sites, where clays, silt, and sand dominate. However, problems can occur when applying the SBCs to offshore sites, where the marine soils may be decomposed from rocks or mixed with artificial fills. This study evaluates the accuracy of six CPTu-based SBCs for marine soils at a site offshore Hong Kong based on 16 CPTu soundings with 25 367 data points by comparing them with composition-based SBCs from borehole records in the vicinity of each sounding. The soil types are determined from six common CPTu-based SBCs. The interpretation of CPTu data is first performed to generate soil type variables comparable to borehole data, followed by a cross-validation study. The soil classification performance of each SBC is quantified by the weighted kappa coefficient and the Kendall correlation coefficient between the soil types generated by the CPTu-based and composition-based SBCs. The classification accuracy for each soil type is also evaluated via the root mean squared error and the mean absolute error. The classified soil types from the CPTu data are associated with a median degree of consistency, indicating the need for combining CPTu-based and composition-based SBCs for marine soil classification.