Stability Evaluation of Slopes Containing a Weak Layer Using Artificial Neural Networks

Man-made slopes including landfills and waste dumps may be situated above a weak layer, the stability of which needs to be evaluated carefully. In this paper, a novel method based on the artificial neural network is proposed to predict the factor of safety for slopes containing a weak layer. A dataset consisting of 50 samples obtained from numerical simulations is used for training and testing. Inputs are selected to be factors influencing the slope stability including not only slope geometry and soil properties but also the properties and thickness of the weak layer. High performance of the trained artificial neural network model is observed from the high correlation coefficient between the predicted and input factor of safety. The trained model is then applied to predict the factor of safety of 144 slopes and compared with some new numerical results, further demonstrating its applicability and reliability.