USING NEURAL NETWORKS TO ANALYZE ROCK SLOPE STABILITY

Abstract

This article examines the feasibility of using artificial neural networks (ANNs) in the geotechnical problem of slope stability prediction by analyzing the accuracy and flexibility of the method when using real datasets. Using a three-layer perceptron with the MLP8-m-1 architecture, neural network predictions are compared with standard engineering methods (Bishop, Janbu, and Spencer), demonstrating high convergence and determination coefficient. A sensitivity analysis of the neural network allowed us to rank the dominant parameters, identifying slope steepness, internal friction angle, and soil moisture content as the factors that have the greatest impact on the stability of objects.