When rotor rubbing occurs, the vibration signal comprises a periodic signal, a transient impact signal and noise. The main components of the periodic signal are the rotating frequency and harmonics thereof. The transient impact signal includes the rotor fault information. According to the characteristics of the local rub-impact fault in a rotor system, an adaptive local iterative filtering (ALIF) method is applied to fault diagnosis of the rotor local rub-impact. The ALIF method is used to decompose rotor vibration signals and can separate the rub, background and noise signals. The fault features of rotor local rub-impact can be extracted from the rotor vibration signal. A case study showed that the ALIF method can be effectively applied to the fault diagnosis of rotor local rub-impact.