An LLM-Driven Framework for Automated Digital Twin Modeling of Manufacturing Cell

At present, digital twin technology has attracted widespread attention as a key approach for enabling intelligent perception and efficient operation and maintenance of equipment. To address the high costs and low efficiency associated with traditional modeling methods, this paper proposes an automated digital twin modeling framework for manufacturing cell that integrates large vision models with a semantic enhancement mechanism. The proposed method reconstructs subunit structures through image perception, injects semantic attributes via large language model (LLM), and achieves system-level scene integration and control. Case studies show that this approach offers significant advantages in modeling efficiency, resource reuse, and system fidelity, providing a novel solution for equipment fault diagnosis and remote operation and maintenance in intelligent manufacturing. Experiments on a representative manufacturing cell demonstrate that the proposed method reduces modeling time by approximately 81.1% compared with manual modeling, while maintaining comparable structural accuracy and semantic consistency. These results highlight the effectiveness and significance of the proposed framework for practical deployment in intelligent manufacturing.