<p>Active dry friction dampers (ADFD) could effectively suppress rotor vibration via nonlinear dry friction effects. Previous studies demonstrated that an optimum normal force exists for the ADFD to minimize rotor vibrations. However, this optimum normal force is significantly influenced by rotational speeds or unbalanced forces, making one optimum design probably not effective in off-design conditions. To address this problem, a nonlinear model predictive control (MPC) algorithm is developed to execute online optimization of normal forces. The proposed MPC utilizes a specially designed augmented Kalman filter to estimate residual unbalances during operation. Based on the estimated residual unbalances, a rotation-speed-dependent prediction horizon is proposed to perform online normal force optimization with limited computational resources. Finally, an iterative algorithm is introduced to solve the nonlinear optimal control problem defined in the rotation-speed-dependent prediction horizon. Numerical and experimental investigations are performed to demonstrate the advantages of the proposed MPC. It is shown that the proposed MPC could adaptively increase the ADFD’s normal forces if the rotor responses are too large while keeping the ADFD’s normal force at a lower level when the rotor responses are relatively small. This makes the ADFD’s vibration suppression performance relatively robust to changes in rotational speeds or unbalanced forces.</p>