Abstract. Yarn tension controlling at heart is tuning servo-motor speed on a restricted range. In this paper, we proposed an advanced model predictive control strategy to solving filament winding system time-delay problem. The application of controlled auto-regressive integrated moving-average (CARIMA) model was described the dynamic time-varying character. The parameters of CARIMA model were solved by recursive least squares parameter estimation method through input-output data. And using the receding optimization method acquire controlling amount in the future. The result of simulation shows that the method is effective and the control structure is simple.