Predicting the lateral capacity of short step-tapered and straight piles in cohesionless soils using an FE-AI hybrid technique
Abstract
Offshore pile foundations are frequently subjected to significant lateral loads, often requiring large-diameter piles. Step-tapered piles have emerged as a cost-effective alternative, offering enhanced lateral capacity with reduced material use. However, reliable and straightforward methods for estimating their lateral bearing capacity remain limited. This study presents a hybrid approach combining three-dimensional finite element (FE) modeling and multi-objective genetic algorithm-based evolutionary polynomial regression (EPR-MOGA) to predict the lateral capacity of short straight and step-tapered piles in cohesionless soils. A parametric study using PLAXIS 3D simulated 580 different pile cases under service-level lateral loads. The mechanisms governing the performance of step-tapered piles were examined and discussed. The FE simulation results were then used to train an artificial intelligence (AI)-based model that produces predictive equations, accurately replicating the FE outputs at a horizontal deflection of 12.5 mm while reducing computational time significantly. The study predictions were compared against the Broms' method, the Characteristic Load Method (CLM), and full-scale field test data. The developed equations account for key geometric and soil parameters, offering a practical and efficient tool for the preliminary design of laterally loaded short piles.