To solve strip deviation in non-woven making process，realize deviation Control of 4 strips，reduce rejection rate，improve the control precision and the productivity. According to the operation characteristics of a non-woven bag making machine in a factory，an improved NSGA-II based model predictive Control（MPC） algorithm was used in strip deviation controlling. The support vector machine (SVM) model was used as the predictive model of controlled object aiming to the nonlinear characteristics of this system. Meanwhile，an improved NSGA-II method was used as rolling optimization strategy of MPC, because of the model's nonlinear characteristics and basic genetic algorithm's slow convergence rate and low precision. Finally, the designed predictive controller was applied in the actual system. It could meet the requirement of real-time and stable operation of the system, the deviation error was basically controlled within ±0.3mm while fast running, and the rate of rejects was reduced from 0.80% to 0.40%. The improved NSGA-II based MPC method used on strip deviation control system of non-woven bag making machine was characteristic of little deviation error, low reject rate, strong reliability，and increased productivity.