Probabilistic assignment of effective friction angles of sands and silty sands from CPT using quantile regression

This paper develops new probabilistic correlations for assigning design values of effective friction angle (ϕ') of clean to silty sands from cone penetration test (CPT) data. New transformation models are proposed using a high-quality database of undisturbed sands. The sand samples have been collected using special methods (freezing, gel samplers, Mazier tube) and tested in laboratory triaxial compression mode. The CPT soundings have been performed using electronic piezocone penetrometers (qt, fs, u2) advanced at the standard rate of 20 mm/s. Two CPT normalization schemes (qt1, Qtn) are investigated. Correlations are obtained through quantile regression, a statistical method which allows the assignment of regression functions which are directly and quantitatively associated with probabilities of exceedance. The derived relationships allow the geoengineer to estimate effective friction angle at any desired level of conservatism. The paper concludes by demonstrating how to apply the results practically in geotechnical design.