Abstract: Based on the observed data of the field simulated rainfall experiment for slope farmland in Loess Plateau of China, the paper used method of artificial neural network model, and established back-propagation network model for slope soil infiltration in different tillage measures (contour tillage, artificial digging, artificial hoeing, linear slope). The network model was trained and predicted by using the observed data of the field simulated rainfall experiment. The results showed that back-propagation network model in this paper were reasonable and can be referred as an effective method for studying soil infiltration laws in slope farmland.