Application of an evolutionary-based approach in evaluating pile bearing capacity using CPT results

Predicting ultimate axial bearing capacity of pile foundations is an important and complicated problem in geotechnical engineering. Cone penetration test (CPT) is a reliable insitu test widely used in the analysis and design of pile foundations. In this study, new CPT-based axial pile bearing capacity models are presented for both cohesionless and cohesive soils using evolutionary polynomial regression (EPR), a branch of evolutionary approaches. A relatively comprehensive database is gathered and divided into training and testing sub-sets to avoid over-fitting. This database includes both coarse and fine grain soils, cone tip resistance and sleeve friction of CPTs, geometry and bearing capacity of piles. The presented models are compared to some previously published ones and their preferences are demonstrated statistically and probabilistically. Proper applicability of the models in predicting axial pile bearing capacity is then confirmed by field verification, compared to analytical and empirical models available in the literature.