Abstract: The deformation and failure of landslides is a complex dynamic process. Through field deformation monitoring of K62 landslide of approach road of A’hai Hydropower Station, a large number of ground surface deformation data are collected, and the deformation laws of the landslide are analyzed. The results show that under the effects of topographical, geomorphological and geological conditions, the rate of surface deformation is asynchronous. The landslide deformation is larger in the upper and lower parts than that in the middle part, and the deformation rates of the landslide are up to 4.06, 0.18 and 48.6 mm/d in the upper, middle and lower parts, respectively. The deformation and failure mechanisms are as follows: drawing of the leading edge pushing of the trailing edge and retaining wall locking of the middle part of the landslide. By means of the uniform design method and artificial neural network based on the measured deformation, a RBF neural network model for Back analysis of the soil parameters of the landslide and slip zone is established. The mechanical parameters are obtained. The FLAC^3D numerical simulation method is adopted to simulate the stress field, distribution of displacement and plastic zone of K62 landslide during dynamic deformation process. The simulated results are consistent with the monitoring ones and provide the design reference for the landslide.