Abstract: This article proposed one kind of new nerve network model,on basis of the thought of the level division and the module,namely RBF mixture neural network model.Successively utilizes the way of reduce－gathers,OLS and the maximum likelihood method in order to solve the problem of modeling and the fast training for big sample collection simplification and enhance the output precision of mixture neural network.The test result indicated the fitting precision of freeform surface was enhanced distinctly by this network model.