부산대학교 도서관

Based on BP neural network, this paper developed an efficient forecasting model for prediction of safety factor of slope stability. Taking unit weight, cohesion, angle of friction, angle and height of slope and void pressure ration as input variables, and safety factor of slope stability as output variable, the prediction model of 6×3×1 BP neural network structure was established. It was found that, the relative error of fitting value of safety factor compared with the observed value for 45 groups of independent variables training BP neural network model was −4.21231% ∼ 2.905645%, the absolute value of the relative error was 0.51893%; And the relative error of predicting value of safety factor compared with the observed value for 6 groups of independent variables validating BP neural network model was −4.8895%∼11.35164%, the absolute value of the relative error was 5.34869%. The following conclusion can be drawn that, the BP neural network prediction model for safety factor of slope stability has good performance and is feasible.