Aiming at the problem of missing damage information caused by the unclear physical meaning of model parameters in the existing structural damage identification method, a damage identification method based on time series model is proposed. Firstly, the general expression of auto-regressive model with eXogenous input (ARX) is derived. The ARX model considering structural dynamic characteristics is established by the simultaneous multi-degree-of-freedom system motion equation. Subsequently, the model is used to predict the nodal acceleration under undamaged conditions, and the parameters of the structural damage, i.e. the damage factor, are constructed according to the difference between the measured data and the predicted data. Finally, the structural damage state is evaluated based on the magnitude and distribution of the damage factor. The numerical results show that the proposed method can perfectly identify single-position and multi-position damage under less measurement dates, and can make a more accurate judgment on the damage degree. At the same time, the influence of the excitation position and measurement noise is small.