Considering the nonlinearity and uncertainty of the high-speed flexible parallel mechanism, a robust model predictive control strategy is presented to suppress vibration response of the system. The actuator is piezoceramic patch, the sensor is strain gauge. The finite element method and modal truncation technique are applied to obtain the dynamic model of the mechanism. The nonlinear, coupling factors and high-order modal effects existed in the dynamic model of the system are considered as disturbance. The modal force is treated as uncertain disturbance, and the measurement noise is considered, the predictive model of the system dynamic response is obtained, and the future output is obtained from the model. The Kalman filter estimator is designed to estimate the state variables of the system. A standard quadratic programming optimization problem is formed where the performance index function minimizes a quadratic performance function. The constraints are the control input voltage and its change rate. The optimization problem is solved to obtain the optimal control output voltages. The rolling optimization control system is constructed to suppress the vibration response of the system. The optimum placements of the actuators and sensors are determined by using the controllability index featured as actuating energy and the observability index featured as signal energy. With a novel 2-DoF parallel mechanism as an example, the experimental modal method is applied to obtain the first two natural frequency and damping ratio. And comparing with the results obtained by the finite element method, it shows that the model is not accurate. Using dSPACE real-time simulation system and MATLAB/Simulink, the control system is built based on the model and the vibration control experimental study is carried out. The experimental results showed the proposed controller can effectively suppress the vibration response, and effectiveness and robustness of the controller is verified.