Abstract: In this paper, an AGC-ASC system of fuzzy neural network based on immune algorithm(IA) is designed. After using IA to optimize the global parameters of membership functions, the local parameters of network are on-line controlled and optimized by gradient-reduced method. Simulation experiments on AGC-ASC system of the two-stand reversible tandem cold mill show that this system has good ability of convergence and resisting disturbance.