A New Optimization-Based Intelligent Method For Evaluation Of Road Embankment Slope Stability

Abstract

The study of slope stability is a common issue for geotechnical and geological engineers. But, suitable computer codes are not frequently user friendly, and supplementary resources accomplished of providing facts for practicing engineers. This study established a genetic algorithm based on artificial neural network (ANN) to predict the stability of embankment soil slope. To provide a neural network training dataset, the total height (H), slope height (h), soil unit weight (γ), slope angle (β), soil friction angle (ϕ), soil cohesion (c), and soil Young’s modulus (E) and their corresponding FS of more than 400 soil slopes were collected from previous publications. Root mean square error (RMSE) and correlation coefficient (R²) were used to assess the generated model's performance and predictive ability. The optimization model with the RMSE value of 0.023 and the R² value of 0.984 is a dependable, straightforward, and valid computational model for estimating the FS of slope, according to the obtained results. Furthermore, the created ANN model is applied to a soil slope case study, and the outcomes show that the suggested model may outperform current approaches and offer superior optimal solutions.