Empirical formulas to predict the axial capacity of driven piles using in-situ dynamic load testing data

Abstract

Accurate estimation of the axial capacity of driven piles is crucial and leads to more economical design. In this research, Artificial neural networks (ANNs) and regression analysis (RA) have been used to develop empirical formulas to estimate piles' axial capacity using experimental dynamic load testing data. The RA model was further improved using a nonlinear fitting software "Eureqa", based on artificial programming. Three input variables were used to build the formulas; these variables are pile cross-sectional area, friction angle at pile tip, and effective stress at pile tip. Simplified mathematical expressions are presented to be used as empirical formulas to estimate piles' axial capacity. An optimize formula was developed using ANNs; it has a coefficient of determination (R 2 ) of 0.92, root mean square error (RMSE) of 478.11, and mean absolute error (MAE) of 366.67. On the other hand, the improved RA formula has R2 of 0.86, RMSE of 450.1, and MAE of 416.91. The results indicated that the predictions of ANNs formula are in close agreement with the experimental data.