To meet the requirements for rapid construction and flexible configuration of digital twin models for discrete manufacturing systems, a construction method for a digital twin mechanism model library is proposed to address the problems of inconsistent representation, poor dynamic adaptability, and low reuse efficiency of existing models. First, a representation paradigm covering multi-dimensional information regarding geometry, physics, behavior, rules, and interfaces is defined. The Modelica meta-model is utilized to achieve the structured representation of heterogeneous information, and an organizational integration mechanism for all-element and multi-level models is designed. Second, a standardized encapsulation method based on the Functional Mock-up Interface (FMI) standard is proposed to define configurable parameters and updatable variables, realizing model instantiation driven by configuration files and parameter updates driven by real-time data. On this basis, a hierarchical model library platform is designed and developed, providing unified support for the storage, management, and application of encapsulated model files and related data. Finally, multi-level verification is conducted using the Space Vector Pulse Width Modulation (SVPWM) component, the spindle subsystem, and a typical discrete manufacturing production line. The results show that the proposed platform and key technologies can improve the organization, management, and reuse of mechanism models, thereby supporting the rapid construction of digital twins for discrete manufacturing systems.