Prediction of Pile Setup for Driven Pipe Piles in Fine-Grained Soils Using Gene Expression Programming

This paper presents the development of a novel model for predicting setup of closed-ended steel pipe (CEP) piles driven in predominantly fine-grained soils using gene expression programming (GEP). The proposed GEP model incorporates both pile and soil properties and is based on dynamic pile load test data obtained from 109 piles installed at 59 different sites. Multiple variable regression analyses conducted on the compiled dataset showed that the most influential parameters in predicting the time-dependent resistance of driven piles were the resistance mobilized at the end of pile installation, time elapsed after installation, pile shaft surface area, and average silt content along the pile length. The data were divided into a training set for model calibration and a validation set for model verification. Sensitivity analysis was performed to assess the model’s robustness. A comparison of the new GEP model with existing pile setup models in the literature was carried out using the collected data, which demonstrated that the proposed GEP model significantly outperformed the tested models. Additionally, data from 22 additional CEP piles were compiled from the literature for model verification purposes. The results showed that the proposed GEP model can predict the total pile resistance with good accuracy.