The blisk is a key component of aero engines, and its surface processing quality directly affects the service performance of the aero engine. In using the new open belt grinding process, different degrees of vibration will occur, and they seriously affect the dimensional accuracy and surface quality of the blisk-blade profile. In view of the above-mentioned problems, this paper first built the physical model for the blisk-blade abrasive belt grinding and analyzed the weak stiffness characteristics of the blisk-blade abrasive belt grinding system. Then, based on the kinetic analysis method, the grinding mechanism of the abrasive belt of the blisk-blade was studied, and the process parameters and stability grinding conditions affecting the grinding stability of the blisk-blade belt were analyzed. In addition, the grinding process of the blisk-blade profile belt was simulated by a numerical method. The influence of different process parameters on the grinding vibration of the blisk-blade was quantified, and the optimal combination of process parameters was obtained. Finally, the feasibility of process parameter optimization was verified experimentally. Studies have shown that the weak stiffness of the blisk-blade grinding system always runs through the entire grinding process, but the impact size has a primary and secondary order. The grinding vibration of the blisk-blade belt is related to the process parameters, such as the grinding positive pressure, the belt speed, the feed rate, and the contact wheel hardness. The primary and secondary orders of each influencing factor are grinding positive pressure, belt speed, feed rate, and contact wheel hardness. The optimal combination of various factors was obtained: belt speed $v_{b} = 3$ m/s, grinding positive pressure $F_{n}=4$ N, feed rate $v_{g}=0.6$ m/s, and contact wheel hardness Hs/A55. Through the sand belt grinding verification experiment on the blisk-blade, the grinding belt grinding quality of the blisk-blade surface roughness and blade profile precision is improved after the process optimization.