Machine Learning Predictive Models for Pile Drivability: An Evaluation of Random Forest Regression and Multivariate Adaptive Regression Splines

Abstract

In recent years, the application of Machine learning (ML) in wide range of industries has grown rapidly. The increase in machine computing capabilities over the past decades has provided the possibility to perform advanced analyses such as ML on big data, which can be effectively used in geotechnical engineering applications. For driven piles, the impact of the piling hammer induces compression and tension stresses in the piles. Hence, an important design consideration is to check that the strength of the pile is sufficient to resist the stresses caused by the impact of the pile hammer. Due to the intrinsic complexity as well as various design variables, pile drivability lacks a precise analytical solution with regard to the phenomena involved. In this paper, the random forest regression (RFR) and multivariate adaptive regression splines (MARS) models are developed for assessing pile drivability in relation to the prediction of the Maximum compressive stresses (MCS) and Blow per foot (BPF). In this study, the 10-fold cross validation method and Lasso regularization is adopted for obtaining the model of superior generalization ability and better persuasiveness results. A database of more than four thousand piles is utilized for model development and comparing RFR with MARS model performance from goodness of fit, running time and interpretability.