Based on RBF neural network and fuzzy theory, an intelligent control method, which can effectively overcome disturbance resulting from grinding efficiency and cyclone’s inlet pressure, is proposed. This method that can make grinding concentration and overflow particle size wellproportioned will allow us to improve flotation grade and increase yield, and therefore realize the optimization of grinding and classification process. The present method is of simple analysis, less time of network learning and training. And high learning precision is high. The simulations show that our approach can also be applied to the control systems that are difficult to build accurate math model.