Innovative FE-AI modelling of lateral resistance of long step-tapered and straight piles in cohesionless soils

Step-tapered piles provide a cost-effective alternative to large-diameter piles for supporting transportation infrastructure subjected to significant lateral forces. However, the literature currently lacks a straightforward method for estimating the lateral bearing capacity of these specialized piles. The present study aims to equip designers with practical and reliable predictive models for the lateral capacity of long step-tapered piles in cohesionless soils. To this end, a comprehensive 3D finite element (FE) analysis was conducted to investigate the parameters influencing the performance of long step-tapered piles and to generate an extensive database suitable for evolutionary polynomial regression (EPR) modelling. The FE results indicated that the optimal design of step-tapered piles involves enlarging the cross-section over a length ( L emb ) corresponding to the depth of the apparent plastic hinge, i.e., the location of the maximum bending moment. Based on the parametric study encompassing 870 cases, two robust predictive models were developed, one for straight piles and one for step-tapered piles, using multi-objective genetic algorithm-based evolutionary polynomial regression (EPR-MOGA). The proposed models incorporate factors related to pile geometry, bending stiffness, and the complex behaviour of soil under lateral loading. Their effectiveness was validated against field measurements for straight piles and FE results for step-tapered piles.