MSC-1DCNN based homogeneous slope stability state prediction method integrated with empirical information

Abstract

Abstract The mechanism of prediction of slope stability is formulated based on its material, geometrical and environmental situation, and the prediction of slope stability has been accepted as a tool for analyzing and predicting future structure stability based on geotechnical properties and failure mechanism. However, the study of slope instability is complex, which is usually difficult to be explained by mathematical methods. The number of slope cases limits the accuracy of slope stability prediction, and the soil or rock parameters of slope are variable, which poses a new challenge for prediction using traditional algorithms. To improve the accuracy of slope stability state prediction, this paper proposes an efficient slope stability state prediction method with the approach of a great robust convolutional neural network named the multi-scale multi-core one-dimensional convolutional neural network (MSC-1DCNN) and substantial Empirical information collected worldwide. Meanwhile, the collected dataset is amplified. Additionally, the probability of failure is calculated by considering the variability of soil or rock parameters. Compared with some state-of-the-art prediction methods, the MSC-1DCNN presents high prediction accuracy. Meanwhile, the proposed method is applied on a slope case, which indicates that our study provides a reliable slope stability state prediction method for homogeneity slope around the world.