Dual statistical approach for correlation of shear wave velocity with the SCPTu parameters

Abstract

Over 30 years of research in the field of shear wave velocity (Vₛ) estimation, scores of researchers were successful in modeling Vs using the Seismic Cone Penetration Test (SCPT) parameters. Through this research work, by utilizing two different statistical approaches in modeling Vₛ from SCPT mechanical parameters, high performing models were produced. This work was done for predicting Vₛ for Kirkland Series soil using the different statistical approaches. One statistical approach was based on stepwise model selection functioning whereas the other one was emulating the Kolmogorov-Gabor polynomial series. The SCPT data was obtained from a standalone Oklahoma Department of Transportation (ODOT) funded project which spanned from February to August 2021. The parameters used for modeling Vₛ were depth, effective overburden stress, sleeve friction and uncorrected cone – tip resistance. The data processing of SCPT parameters consisted of 29 Vₛ profiles of datasets along with datasets of selected parameters, which were averaged to set up modeling process for averaged datasets of all parameters. All four variables were found to be positively correlated to one another. Sleeve friction was noted to be most positive correlated to Vₛ. After the extensive modeling process, models M15 as well as stepwise model yielded R² values of 0.9911 and 0.9621. These models were found to be optimal models produced statistically following the modeling process and were compared with select literary model functions. Additionally, the models of literature resulted in higher RMSE values whereas RMSE values as well as residual error plot showed lower values and variance.