Estimation on SPT number based on the prediction N-value model for soft soils / Raudhah Razali ... [et al.]

This paper proposes a precise multiple linear regression model for N-Value to estimate the Standard Penetration Test (SPT) number for soft soil. The study used 76 data from soil properties for N-Value model development collected from site investigation report. Meanwhile, 47 datasets for N-Value model validation collected from other site investigation report and technical paper. The N-value model developed in this study has met the requirements and has been verified according to the statistical validity requirements. The results shows that the developed N-Value model has the significant accuracy linear regression with R2 = 68%. For the model accuracy it shows the small deviation discrepancy results for Mean Square Error (MSE) is 0.4, Mean Absolute Error (MAE) is 0.4, and Mean Absolute Percentage Error (MAPE) is 0.31 with a determination coefficient, R2 = 93.7%. Accordingly, the N-Value models developed in this study demonstrate a good prediction capability for SPT number. The proposed model will serve as a reference for civil engineers in estimating SPT number for provide information on the geotechnical engineering properties of soil, especially for the determination of bearing capacity and settlement.