Predicting Standard Penetration Test (SPT) N-Values Using Drilling Response Measurements Collected during the Installation of Rigid Inclusions

This paper evaluates the performance of supervised machine learning (ML) algorithms in predicting standard penetration test (SPT) N-values. The study utilized drilling response measurements (DRM) collected from the installation of rigid inclusions at a highly heterogeneous municipal solid waste (MSW) site. A range of ML algorithms were evaluated in this paper, including linear regression, decision trees, random forests, support vector machines, gradient boosting, and artificial neural networks. The ML algorithms were trained and evaluated on DRM data along with SPT boring data available for the project. The results of this study showed that ML algorithms can be used to predict SPT N-values to give rudimentary information about the relative density of the soil from DRM data.