Development of Correlations between SPT-CPT Data for Liquefaction Assessment using R

Abstract

This paper presents a method for selecting and processing the field data for correlation and comparison of standard penetration test – cone penetration test (SPT-CPT). The correlations of SPT-CPT were done using traditional and statistical methods. SPT, CPT field tests were used in conjunction with variety of borehole seismic testing for a number of locations to adapt traditional site research approaches to geotechnical earthquake engineering. The correlations between N60 of SPT and cone tip resistance qc and other parameters of CPT data under liquefaction conditions, were developed using regression modelling. In this paper, the SPT-CPT correlations have been developed using different type of regression methods namely linear regression (LR), locally estimated scatterplot smoothing (LOESS), multivariate adaptive regression splines (MARS) and support vector machine (SVM). Correlation between N60 and qc which was developed using support vector regression (SVR) model is giving 90.53% efficiency. Correlation which was developed between N60 and qc taking other parameters of CPT data such as fines content and mean particle size D50 in SVR model is giving 99.99% efficiency. By using the above theses correlations, SPT N-value may be evaluated using CPT data. Predicted N60 values from these correlations are compared with measured N60 values from existing literature and seismic tests and it was found to be good.