Research on dynamic model construction method based on RecurDyn secondary development.

Efficient multibody dynamics (MBD) modeling is critical for the dynamic analysis of complex engineering equipment, yet conventional manual workflows remain labor-intensive, time-consuming, and error-prone. This paper proposes an automated MBD modeling methodology based on Python-driven meta-programming and automated C# code generation, with the cutting unit of a drum shearer selected as a case study. The proposed method first extracts Computer-Aided Design (CAD) assembly features and then employs a hybrid lexical-semantic dual-stream architecture with a multi-level gating mechanism for robust component identification. To improve fault tolerance under non-standard naming conditions, a pre-trained multilingual Transformer model is incorporated as an auxiliary semantic matching module. Subsequently, a customized execution engine invokes the RecurDyn ProcessNet (PNet) to automatically complete material assignment, topology integration, constraint and drive generation, and gear contact definition. To validate the methodology, kinematic simulations were conducted on three shearer cutting units with different structural complexities. Model reliability was further verified through rigid-flexible coupled Discrete Element Method-Multi Flexible Body Dynamics (DEM-MFBD) simulations and corresponding physical experiments on a shearer testing platform. Results show that the method requires less than 1% of traditional modeling time with high accuracy. The maximum relative error of transmission rotational speeds is only 0.025%. Furthermore, the time-domain Root Mean Square (RMS) error of the drum vibration acceleration between simulation and experiment is 4.8%, while the maximum relative error of the primary characteristic frequencies is 4.1%. This rule-driven methodology provides a transferable framework for the rapid modeling and simulation of diverse complex mechanical MBD systems.