Abstract: To improve the decoupling effectiveness of multiple variables in polyacrylonitrilc carbon fiber (PANCF) production line, based on the artificial cytokine network (ACN) and its mathmatical model, a network structure of the stretching process was proposed and investigated. With decoupling compensation algorithm embedded in the calculation center, a complete cytokine network based decoupling network (CNDN) was proposed, which realized the decoupling of three variables in the stretching tank:temperature, concentration, and liquid level. Simulation results show that the CNDN not only can rapidly respond the set-points of control variables, but also completely eliminate the influence on other variables coupling with it; moreover, it has a better ability of resistance against the interference.