Comparing nonlinear regression analysis and ‎artificial neural networks to predict geotechnical ‎parameters from standard penetration test

Abstract

At the beginning the twenty-first century, a lot of high-level methods have become ‎available in geotechnical engineering in order to deal with the complexity and ‎heterogeneity encountered in soil, Statistical modeling (i.e. regression analysis method) ‎was used to estimate the relationships among two or more variables, however in the early ‎nineties an application of a new system emerged which gave excellent results in solving a ‎lot of problems by learning from the available data so-called ”artificial neural network”.‎ The aim of this study is to apply both methods, nonlinear regression analysis and ‎artificial neural networks in order to predict geotechnical parameters from standard ‎penetration test in all soil’s types; Comparison of the results using correlation’s ‎coefficient (R) and Root Mean Squared Error (RMSE) is done between both methods; ‎About 400 samples, over 65 boreholes in the Algiers area have been collected and were ‎used in this study, The results show the superiority of ANN‎ method in predicting data ‎that seems closer to experimental values compared to NRA method.

Keywords

• artificial neural network
• regression analysis
• standard penetration number and ‎geotechnical parameters