Abstract: Using the data of underground water in Dashiban landslide in 1983, an artificial neural network method is developed to predict the groundwater table in observation holes with both drainage and non-drainage systems. The groundwater table is considered as a fixed water head boundary to simulate the seepage of the landslide. The limit equilibrium analysis and evaluation of the effectiveness of drainage system are made in combination with result of stability analysis of landslide and according to the variation of groundwater related to landslide drainage system. The result shows that the drainage control engineering is reasonable and effective.