Abstract: The slope instability has always been a key technical issue for the safe production in open-pit mine.The realization of the intelligent warning of slope instability is the core of instability research.The early warning model based on RBF neural network was established，taking the high-steep slope engineering of Aoshan pit in Nanshan Iron Mine of Masteel as a case.Gradient descent algorithm for training was improved，and according to the experience，the algorithm's learning step was set up.Six factors of cohesion，internal friction angle，slope angle，slope height，ratio of pore water pressure and bulk density were selected as network input units.And，25 sets of sample data selected were used to complete the learning of RBF neural network.Then，the early warning model was used to make early warning analysis on instability of two slopes of Aoshan pit in Nanshan Iron Mine.Stability classification of the two slopes is respectively level 1 and level 3，that are，extremely stable and basically stable，which are accordant with the current actual situation.This early warning method is worth being applied and spread.

Key words: RBF neural network, Slope instability, High-steep slope, Early warning model