Abstract: To address the issues of low model reconstruction efficiency, weak dynamic response, and poor virtual-real interaction in valve manufacturing processes, a dynamic control method for valve machining quality driven by digital-model fusion is proposed. A four-dimensional digital twin model of valve manufacturing is constructed based on geometry, logic, mechanism, and process, which enables bidirectional mapping between the physical manufacturing process and the virtual model. A machining error compensation model based on step nodes is established, and a nonlinear error compensation method driven by data-model fusion is developed using the dynamic time warping algorithm. Experimental verification is performed on a valve sleeve manufacturing case, and the dimensional accuracy of the part is improved by 44.82% after error compensation. The results indicate that real-time monitoring of valve machining quality and adaptive error compensation are achieved.