A Data-Driven Approach to Predict Shear Wave Velocity from CPTu Measurements: An Update

Abstract

This paper is an extension of our previous research investigating the potential of machine learning models to estimate shear wave velocity (Vs) from piezocone penetration test (CPTu) measurements. The aim of this update is to examine the effect of incorporating geographical information, namely latitude and longitude, as input parameters to the machine learning models. New models are developed by incorporating both CPTu parameters and spatial coordinates as input features and are compared to models developed with only CPTu parameters. Furthermore, SHAP (SHapley Additive exPlanations) analysis is employed to assess the importance of different features and variables in the developed machine learning models. The results show improvement in prediction performance when adding geographical data, indicating the influence of geological variations on Vs. The paper shows the potential of using geospatial information to improve the data-driven approach for estimating soil properties from CPTu tests when large worldwide datasets are available.

Full Paper