Development and Calibration of Empirical and Statistical Models for SPT-N Prediction in Fine Grained Soils

Standard Penetration Number Index Properties Soil Investigations Empirical Method Multiple Regression Analysis

Authors

Downloads

This study aims to develop and calibrate predictive models for the Standard Penetration Test number (N-SPT) in cohesive soils with a Liquid Limit between 20% and 60%. The objectives were to evaluate and compare empirical estimations based on the Consistency Index (CI) against statistical models derived from Multiple Linear Regression (MLR). Methods involved the analysis of a comprehensive dataset containing 469 samples obtained from the Thailand-China High-Speed Railway project and established literature, utilizing soil index properties (Liquid Limit (LL), Plastic Limit (PL), and water content (wn)) alongside unit weight (γ) as independent variables. Findings demonstrate that the MLR model provides significantly higher predictive reliability with a coefficient of determination (R2) of 0.982, compared to the empirical method (R2 = 0.667). To enhance practical application, both models were calibrated using a 90% confidence level modification factor. Novelty/Improvement: This research identifies unit weight as a critical parameter that, when integrated with index properties, substantially improves the accuracy of N-SPT estimations. The resulting framework provides geotechnical engineers with a validated, high-precision tool for soil strength estimation, effectively accelerating soil investigation processes while maintaining high reliability in design parameters.