Predicting Shear Stress Parameters in Consolidated Drained Conditions Using Artificial Intelligence Methods

Abstract

Using the direct shear stress test for estimating shear stress parameters is considered to be of great importance, mainly for enhancing and strengthening soils, assessing their bearing capacity, and predicting potential risks that could bring harm to foundations. However, conducting the test in consolidated drained conditions is quite expensive and time-consuming (e.g., up to three months in consolidated clay). To our knowledge, few researchers have suggested simple models in undrained conditions to experimentally estimate these parameters. However, in large projects and slope studies, testing in consolidated drained conditions is more important because these conditions mimic reality. The current study aims to suggest a new model for estimating shear stress parameters. The reliability of the approach was tested through comparing several models of multiple regression analysis, genetic programming, and artificial neural networks. These models were tested on 98 samples of Algiers soil. The results showed the efficiency of the artificial neural network method with two hidden layers, which provided the best appropriate model, and the most approached results to experimental data, as compared with the other models. Based on these findings, this study proposes a structural flowchart for effectively predicting shear stress parameters effectively in future studies.