Artificial neural network based predicting model for evaluating stability of landslide

摘要

Nine influencing factors of slope stability are selected as input variables, and chose stability coefficient under different conditions are set as output variables, by confirming some other network parameters,a forecasting model based on artificial neural network has been established. 17 landslides are selected as the training and check-up swatches which are located along the same rebuild-road as Shifo-temple landslide.During the training process, by contrasting the results that be trained by different functions, it’s found that the ‘traingdx’ function which using plus-momentum and self-adapting arithmetic, can avoid the problem of local-infinitesimal-value and adjust the study speed to improve the training effectiveness. Lastly, the article forecasts the stability of Shifo-temple Ⅱ landslide with the established neural network model; the result shows that the artificial neural network has well validity and nicer foreground in the process of landslide stability forecasting.