Considering the spatial variability and the updating of parameters of soil strength,a Convolutional Neural Networks(CNN)model was employed to analyze slope reliability.Taking the shear strength parameters as random varia-bles,the correlated Gaussian random fields were simulated using the Cholesky decomposition midpoint method.A sample database was obtained with finite element analysis.The CNN model was trained and validated to predict the safety factor of the slope,and then the slope reliability was calculated.Based on newly added information/observation samples,Bayes-ian method was used to update the shear strength parameters and random fields,the aforementioned CNN model was ap-plied to evaluate the slope reliability.Taking two slopes as examples,the results show that the CNN model has strong learning and generalization ability,and it can significantly improve the reliability calculation efficiency of the anisotropic random field slope,the model also has good predictive performance after the parameters are updated.Parameters updating that integrates on-site observation information can provide more realistic reliability evaluation.
NETL
NSTL
万方数据
