The large number of aero-engine rotor components and the large computational volume of high-dimensional complex models lead to difficult dynamic analysis and long computation times， which are disadvantageous to the efficiency of rotor structure design and dynamics verification. Based on the component modal synthesis method， a novel multi-stage modal reduction strategy is proposed for the modal reduction of a large complex system with many components. The internal freedom degrees of each substructure are reduced in parallel using fixed interface modal reduction， while the couplings between the substructures are retained completely. By defining a new level of substructure through substructure combination， the multi-stage modal reduction is applied to an additional reduction， and the hybrid mode synthesis is subsequently combined to construct the branch mode and significantly re? duce the dimensionality of the rotor FEM model. Meanwhile， the dynamic characteristics of key substructures and the key dynamic characteristics of the vibration system are preserved. This computational strategy is used to establish a low-dimensional reduced model of a missile engine rotor system， and the reduced model is used to improve the efficiency of rotor dynamics analysis and ac? celerate the design optimization of bearing stiffness parameters. The results show that the time required for rotor dynamics analysis is reduced by 99.5%， and the accuracy error does not exceed 0.1% compared with ANSYS calculations. The computational strate? gy can be used for rapid analysis of multi-component high-dimensional complex systems.