ABSTRACT
The use of machine learning modelling to predict shear wave velocity (VS) from piezocone penetration tests (CPTu) is presented. A large dataset of paired VS-CPTu data (n = 104,054) compiled from seismic piezocone (SCPTu) soundings completed in a wide variety of soil types with various stress histories and geological environments was used to develop machine learning models to directly estimate VS from CPTu data. The impact of soil microstructure on the results was investigated and separate models were developed to predict VS in cemented and uncemented soils. The results of machine learning models outperformed the existing widely used CPT-based relationships to predict VS.
