Abstract: Aiming at the industrial pain points of complex thin-walled die-cast components in 5-axis machining, such as difficulty in pose estimation and low machining accuracy caused by poor blank consistency and lack of physical positioning datums, a 5-axis digital self-alignment machining method based on on-machine measurement is proposed. Firstly, the workpiece pose determination problem is abstracted into a six-degree-of-freedom spatial transformation mathematical model. The RANSAC algorithm is employed to eliminate noise on the blank surface, and the singular value decomposition algorithm is integrated to achieve precise registration between the measured point clouds and the theoretical model. Secondly, addressing the error propagation of rotary axes in 5-axis machining, a process strategy of "coarse-to-fine hierarchical alignment and dynamic datum correction" is proposed to realize the mathematical reconstruction and automatic compensation of the machining datum through closed-loop feedback. Finally, a small-batch production trial of 400 automotive throttle bodies was conducted. The results indicate that the proposed method reduces and stabilizes the coaxiality deviation of key hole systems from an average of 0.012 mm to within 0.002 mm. Furthermore, the process capability index is improved from 0.73 to 1.67, and the comprehensive production efficiency is increased by 20%.