A Min-Max structure combined with several linear limit controllers are utilized for traditional aero-engine safety control, which is complex and conservative, therefore a multivariable control system based on self-tuning model predictive control method is proposed that owns the ability to handle constraints directly. The control system consists of two parts. One is called prediction model part and the other is called control strategy part. In the prediction model part, firstly, the nonlinear component-level model of the aero-engine is converted into a linear state space expression. Secondly, the state space expression will be combined with the simplified actuator models so that the prediction model can partly reflect the effects of the actuators. What’s more, the process of the linearization is repeated during every control period, so that the control system will renew the prediction model at different operation points in the flight envelop. The control strategy part computes the control variables online along with the aero-engine feedback data, which are determined by solving a quadratic programming problem with constraints in finite time domain. In this way, the calculation burden of linear constrained optimization is less than that of a nonlinear optimization problem, so the proposed method is suitable for the application in aero-engine. Considering the effects of mismatch situation that exists between the prediction model and actual aero-engine, the feedback correction is introduced in the system. In this paper, a turbofan gas turbine engine with an afterburner is chosen to verify the effectiveness of the multivariable control system. The results of the simulation show that the aero-engine can operate well at different working conditions.