Stability Prediction of Rock Slope Based on Fuzzy Clustering GA-FNN Model

The stability evaluation of rock slope is an important task of geotechnical engineering, which needs taking account of various factors, such as topography, lithology and geological structure. Moreover, the slope is a complex open system with uncertainty, nonlinearity and fuzziness, thus the traditional analysis method is difficult to evaluate complicated slopes stability conforming to reality. The combination of fuzzy clustering and neural network prediction can provide a solution for this problem. Tacking slope topography and rock parameters of Hoek–Brown criterion widely applied in rock engineering as the main influencing factors of slope stability, namely slope height, slope angle, geological strength index (GSI), rock unit weight γ, disturbance factor D, uniaxial compressive strength σci and rock empirical parameter mi. Due to the complexity and diversity of slope lithology and rock structure, and slope shape, the wide input data of neural network seriously affects the prediction accuracy. Therefore, fuzzy clustering is applied to classify and preprocess the training data. Then the classified fussy data is delivered to the Fuzzy Neural Network (FNN) to train the model and predict the stability state and safety factor of the slope. Besides, the Genetic Algorithm (GA) is used to optimize the weight and threshold of FNN network by crossover and mutation, and GA-FNN prediction model is establish. 200 slope samples are collected from the slope examples of the references and the expansion examples by Slide simulation. The results show that the prediction model based on fuzzy clustering GA-FNN is suitable for the stability prediction of rock slope with high speed and high accuracy.