Abstract
The fundamentals of artificial neural networks (ANN) are implemented to develop a prediction model to simulate the mechanism of fully softened shear strength (FSS). The model is generated from a laboratory database containing 201 laboratory strength tests with FSS secant friction angle values as the output and soil properties as the inputs. The objective of the study is to detect the inputs from the parameter space that controls the response in a hierarchical manner. From the results, a variable assessment indicates that among the studied factors, the main factors affecting the FSS secant friction angle are the plasticity index, the clay fraction, and the normal effective stress. Predictor profilers and surfaces are provided as a visual way of studying how changes in the factors influence the response—the effects of the plasticity index in the FSS secant friction angle decrease after about 100% plasticity index. The clay fraction profiler shows that an increase in the clay fraction decreases the FSS secant friction angle at a constant rate. Furthermore, the strong influence from normal stress in the FSS secant friction angle decreases after a cutoff value of around 150 kPa, which agrees with the curvature of the failure envelope. Insights regarding the interaction and correlation coefficients of the studied factors with the FSS secant friction angle are given.
Original language | English |
---|---|
Title of host publication | Proceedings of the technical sessions of the international foundations congress & equipment expo ASCE, Dallas, Texas, May 10-14, 2021. |
DOIs | |
Publication status | Published - 2021 |