An Artificial Neural Network Method for Forecasting the Stability of Soil Slopes

Abstract

Artificial neural network (ANN) methods, based on sophisticated models, have been developed recently that can predict slope stability. In this study, we have developed a genetic algorithm (GA) based on ANN to assess the stability of soil slope. Firstly, an ANN-based genetic algorithm was trained for nonlinear input-output mapping of the slope. A total of 190 soil slopes with unique values of shear strength properties (friction angle, cohesion, and unit weight), geometric parameters (slope angle and slope height), and corresponding factor of safety (FS) have been collected to give a neural network training dataset. Then, a three-layer neural network model is established based on GA. The prediction and performance ability of the established model is assessed using the correlation coefficient (R2). By the outcomes, the trained ANN model with the R2 value of 0.98 is reliable, valid, and simple for evaluating the soil slope stability and estimating the FS. Additionally, the proposed neural network model is applied to a case of soil slope from prior studies. Findings show that the developed ANN model can be versatile in studying the stability of soil slopes.