An Effective Artificial Intelligence Approach for Slope Stability Evaluation

Abstract

In this study, an effective intelligent system based on artificial neural networks (ANN) and a new version of the sine cosine algorithm (SCA) is developed to evaluate and predict the FOS of homogenous slopes under static and dynamic loading. In the first step, an effective hybrid optimization algorithm based on the adaptive sine cosine algorithm (ASCA) and pattern search (PS), namely ASCPS, is proposed and verified using a set of benchmark test functions. Then, the new algorithm, along with the Morgenstern and Price method is applied for seismic slope stability evaluation. To provide a neural network training dataset, a set of 189 slopes with different values of slope height, slope angle, friction angle of soil, soil cohesion, and horizontal acceleration coefficient have been analyzed and their corresponding FOS have been recorded. In the next step, the proposed ASCPS algorithm is implemented for training the ANN model using the collected database. The performance and prediction capacity of the developed model are evaluated using root mean square error (RMSE) and correlation coefficient (R). According to the obtained results, the ANN model with the RMSE value of 0.023 and the R value of 0.984 is a reliable, simple, and valid computational model for estimating the FOS and evaluating the slope stability under static and earthquake loads. In addition, the developed ANN model is applied to a case study of slope stability from previous studies, and the results reveal that the proposed model may provide better optimal solutions and outperform existing methods.

Keywords

• Neural network
• metaheuristic
• hybrid algorithm
• slope stability
• earthquake load